
# Notebook timestamp
from datetime import datetime
dt_string = datetime.now().strftime("%d/%m/%Y %H:%M:%S")
print(f"Updated {dt_string} (GMT)")

Updated 25/03/2023 01:25:12 (GMT)


# NB start time
import time
a = time.time()


!pip install ipython-autotime
%load_ext autotime

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting ipython-autotime
  Downloading ipython_autotime-0.3.1-py2.py3-none-any.whl (6.8 kB)
Requirement already satisfied: ipython in /usr/local/lib/python3.9/dist-packages (from ipython-autotime) (7.9.0)
Requirement already satisfied: prompt-toolkit<2.1.0,>=2.0.0 in /usr/local/lib/python3.9/dist-packages (from ipython->ipython-autotime) (2.0.10)
Collecting jedi>=0.10
  Downloading jedi-0.18.2-py2.py3-none-any.whl (1.6 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.6/1.6 MB 63.9 MB/s eta 0:00:00
Requirement already satisfied: pickleshare in /usr/local/lib/python3.9/dist-packages (from ipython->ipython-autotime) (0.7.5)
Requirement already satisfied: pygments in /usr/local/lib/python3.9/dist-packages (from ipython->ipython-autotime) (2.14.0)
Requirement already satisfied: traitlets>=4.2 in /usr/local/lib/python3.9/dist-packages (from ipython->ipython-autotime) (5.7.1)
Requirement already satisfied: backcall in /usr/local/lib/python3.9/dist-packages (from ipython->ipython-autotime) (0.2.0)
Requirement already satisfied: pexpect in /usr/local/lib/python3.9/dist-packages (from ipython->ipython-autotime) (4.8.0)
Requirement already satisfied: decorator in /usr/local/lib/python3.9/dist-packages (from ipython->ipython-autotime) (4.4.2)
Requirement already satisfied: setuptools>=18.5 in /usr/local/lib/python3.9/dist-packages (from ipython->ipython-autotime) (67.6.0)
Requirement already satisfied: parso<0.9.0,>=0.8.0 in /usr/local/lib/python3.9/dist-packages (from jedi>=0.10->ipython->ipython-autotime) (0.8.3)
Requirement already satisfied: six>=1.9.0 in /usr/local/lib/python3.9/dist-packages (from prompt-toolkit<2.1.0,>=2.0.0->ipython->ipython-autotime) (1.16.0)
Requirement already satisfied: wcwidth in /usr/local/lib/python3.9/dist-packages (from prompt-toolkit<2.1.0,>=2.0.0->ipython->ipython-autotime) (0.2.6)
Requirement already satisfied: ptyprocess>=0.5 in /usr/local/lib/python3.9/dist-packages (from pexpect->ipython->ipython-autotime) (0.7.0)
Installing collected packages: jedi, ipython-autotime
Successfully installed ipython-autotime-0.3.1 jedi-0.18.2
time: 363 µs (started: 2023-03-25 01:25:20 +00:00)


# Import all the relevant libraries needed to complete the analysis, visualization, modeling and presentation
import pandas as pd
import numpy as np
import os

import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
sns.set_style('darkgrid')

from scipy import stats
from scipy.stats import zscore

from sklearn import preprocessing
from sklearn.preprocessing import LabelEncoder
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import MinMaxScaler
from sklearn.preprocessing import RobustScaler
from sklearn import model_selection
from sklearn.model_selection import train_test_split, cross_val_score, GridSearchCV
from sklearn.model_selection import RandomizedSearchCV, StratifiedKFold, KFold
from sklearn.metrics import accuracy_score, confusion_matrix, classification_report 
from sklearn.metrics import ConfusionMatrixDisplay, precision_score, recall_score 
from sklearn.metrics import precision_recall_curve, roc_curve, auc, roc_auc_score
from sklearn.metrics import f1_score
# from sklearn.metrics import plot_precision_recall_curve, average_precision_score


from sklearn.neighbors import KNeighborsClassifier
from sklearn.linear_model import LinearRegression, LogisticRegression, RidgeClassifier, SGDClassifier
from sklearn.naive_bayes import GaussianNB, BernoulliNB, MultinomialNB
from sklearn.svm import SVC

# from sklearn.decomposition import PCA
# from scipy.cluster.hierarchy import dendrogram, linkage
# from scipy.cluster.hierarchy import fcluster
# from sklearn.cluster import KMeans 
# from sklearn.metrics import silhouette_samples, silhouette_score

# import xgboost as xgb
# from xgboost import plot_importance
# from lightgbm import LGBMClassifier

from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import AdaBoostClassifier
from sklearn.ensemble import BaggingClassifier
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.ensemble import VotingClassifier

from imblearn.over_sampling import RandomOverSampler
from imblearn.over_sampling import SMOTENC, SMOTE, ADASYN
from imblearn.under_sampling import RandomUnderSampler

# import re
# from nltk.corpus import stopwords
# from sklearn.feature_extraction.text import CountVectorizer
# from nltk.stem.snowball import SnowballStemmer
# import pandas_profiling as pp

# import gensim
# import logging

# import cv2
# from google.colab.patches import cv2_imshow
# from glob import glob
# import itertools

import tensorflow as tf
from tensorflow.keras.models import Sequential, Model
from tensorflow.keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D, Conv2DTranspose
from tensorflow.keras.layers import Activation, GlobalMaxPool2D, GlobalAveragePooling2D
from tensorflow.keras.layers import UpSampling2D, Input, Concatenate, Reshape
from tensorflow.keras.layers import BatchNormalization, LeakyReLU
from tensorflow.keras.optimizers import Adam, RMSprop, SGD, Adagrad

from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau, ModelCheckpoint
from tensorflow.keras.metrics import Recall, Precision
from tensorflow.keras import backend as K

from tensorflow import keras
from keras.utils.np_utils import to_categorical  
from keras.utils import np_utils
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.wrappers.scikit_learn import KerasClassifier, KerasRegressor

import warnings
warnings.filterwarnings("ignore")

import random
from zipfile import ZipFile

# Import pickle Package
import pickle

# Set random_state
random_state = 42

# pd.set_option('display.max_columns', None)
# pd.set_option('display.max_rows', None)

time: 3.82 s (started: 2023-03-25 01:25:20 +00:00)


# Install pydicom package
!pip install pydicom

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting pydicom
  Downloading pydicom-2.3.1-py3-none-any.whl (2.0 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.0/2.0 MB 65.0 MB/s eta 0:00:00
Installing collected packages: pydicom
Successfully installed pydicom-2.3.1
time: 7.15 s (started: 2023-03-25 01:25:24 +00:00)


# Import other relevant packages
from numpy import asarray

import cv2
import glob
import time
import pydicom as dcm
import skimage
import math

from matplotlib.patches import Rectangle
from skimage import feature, filters
from PIL import Image

from functools import partial
from collections import defaultdict
from joblib import Parallel, delayed
from tqdm import tqdm, tqdm_notebook

# Tensorflow / Keras
from tensorflow.keras.layers import *
from tensorflow.keras import Model
from keras import models
from keras import layers

from tensorflow.keras.applications import MobileNet, MobileNetV2, VGG19, EfficientNetV2B3
from tensorflow.keras.applications.vgg16 import VGG16
from tensorflow.keras.applications.mobilenet import MobileNet
from tensorflow.keras.applications.mobilenet import preprocess_input 
import tensorflow.keras.utils as pltUtil
from tensorflow.keras.utils import Sequence
from tensorflow.keras.applications.resnet import ResNet50
from tensorflow.keras.applications.resnet import preprocess_input as resnetProcess_input

sns.set_style('whitegrid')
np.warnings.filterwarnings('ignore')

time: 672 ms (started: 2023-03-25 01:25:31 +00:00)


from google.colab import drive
drive.mount('/content/drive')

Mounted at /content/drive
time: 33.7 s (started: 2023-03-25 01:25:31 +00:00)


# Current working directory
%cd "/content/drive/MyDrive/MGL/Project-Capstone/"

# # List all the files in a directory
# for dirname, _, filenames in os.walk('path'):
#     for filename in filenames:
#         print(os.path.join(dirname, filename))

/content/drive/MyDrive/MGL/Project-Capstone
time: 116 ms (started: 2023-03-25 01:26:05 +00:00)


# # Path of the data file
# path = 'rsna-pneumonia-detection-challenge.zip'

# # Unzip files in the current directory

# with ZipFile (path,'r') as z:
#   z.extractall() 
# print("Training zip extraction done!")

time: 385 µs (started: 2023-03-25 01:26:05 +00:00)


# List files in the directory
!ls

 Capstone-Project-Grp5-Amit-Kumar       Test
 cr_model.h5			        Test-1
'GCP Credits Request Link - RSNA.txt'   Test-2-Modeling-Interim
 Mask_RCNN			        Test-3-Modeling-Final
 stage_2_detailed_class_info.csv       'Test-4-Hyperparameter Tuning'
 stage_2_sample_submission.csv	        Test-5-UNet
 stage_2_test_images		        train_labels_merged_metadata.csv
 stage_2_train_images		        unet_model.h5
 stage_2_train_labels.csv
time: 1.53 s (started: 2023-03-25 01:26:05 +00:00)


class_info_df = pd.read_csv('stage_2_detailed_class_info.csv')
train_labels_df = pd.read_csv('stage_2_train_labels.csv')

time: 850 ms (started: 2023-03-14 06:17:08 +00:00)


print(f"Detailed class info -  Rows: {class_info_df.shape[0]}, Columns: {class_info_df.shape[1]}")
print(f"Train labels -  Rows: {train_labels_df.shape[0]}, Columns: {train_labels_df.shape[1]}")

Detailed class info -  Rows: 30227, Columns: 2
Train labels -  Rows: 30227, Columns: 6
time: 6.13 ms (started: 2023-03-14 06:17:09 +00:00)


class_info_df.sample(10)

time: 104 ms (started: 2023-03-14 06:17:09 +00:00)


train_labels_df.sample(10)

time: 36.1 ms (started: 2023-03-14 06:17:09 +00:00)


# Understand the missing data points
def missing_data(data):
    total = data.isnull().sum().sort_values(ascending = False)
    percent = (data.isnull().sum()/data.isnull().count()*100).sort_values(ascending = False)
    return np.transpose(pd.concat([total, percent], axis=1, keys=['Total', 'Percent']))

missing_data(train_labels_df)

time: 110 ms (started: 2023-03-14 06:17:09 +00:00)


missing_data(class_info_df)

time: 66.6 ms (started: 2023-03-14 06:17:09 +00:00)


# Duplicate patients in the class_info_df dataset
duplicateClassRowsDF = class_info_df[class_info_df.duplicated(['patientId'])]
duplicateClassRowsDF.shape

## There are 3543 duplicates similar to the train_labels_df dataset

(3543, 2)

time: 15.1 ms (started: 2023-03-14 06:17:09 +00:00)


duplicateClassRowsDF.head(2)

time: 12.2 ms (started: 2023-03-14 06:17:09 +00:00)


## The same patient id has the same class even though they are duplicate
class_info_df[class_info_df.patientId=='00704310-78a8-4b38-8475-49f4573b2dbb']

time: 23.1 ms (started: 2023-03-14 06:17:09 +00:00)


# Duplicate patients in the train_labels_df dataset
duplicateRowsDF = train_labels_df[train_labels_df.duplicated(['patientId'])]
duplicateRowsDF.shape

## There are 3543 duplicates similar to the class_info_df dataset

(3543, 6)

time: 25.2 ms (started: 2023-03-14 06:17:09 +00:00)


duplicateRowsDF.head(2)

time: 18.4 ms (started: 2023-03-14 06:17:09 +00:00)


## Examining one of the patient id which is duplicate , we can see that the x,y, widht and height is not the same
## This indicates that the same patient has two bounding boxes in the same dicom image
train_labels_df[train_labels_df.patientId=='00436515-870c-4b36-a041-de91049b9ab4']

time: 21.3 ms (started: 2023-03-14 06:17:09 +00:00)


# Clear the matplotlib plotting backend
%matplotlib inline
plt.close('all')

time: 6.93 ms (started: 2023-03-14 06:17:09 +00:00)


# Understand the 'class' variable
f,axes=plt.subplots(1,2,figsize=(17,7))
class_info_df['class'].value_counts().plot.pie(autopct='%1.1f%%',ax=axes[0])
sns.countplot('class',data=class_info_df,ax=axes[1])
axes[0].set_title('Pie Chart for class')
axes[1].set_title('Bar Graph for class')
plt.show()

time: 355 ms (started: 2023-03-14 06:17:09 +00:00)


# Let's merge the two datasets using Patient ID as the merge criteria.
train_class_df = train_labels_df.merge(class_info_df, left_on='patientId', right_on='patientId', how='inner')

time: 30.8 ms (started: 2023-03-14 06:17:10 +00:00)


train_class_df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 37629 entries, 0 to 37628
Data columns (total 7 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   patientId  37629 non-null  object 
 1   x          16957 non-null  float64
 2   y          16957 non-null  float64
 3   width      16957 non-null  float64
 4   height     16957 non-null  float64
 5   Target     37629 non-null  int64  
 6   class      37629 non-null  object 
dtypes: float64(4), int64(1), object(2)
memory usage: 2.3+ MB
time: 28.9 ms (started: 2023-03-14 06:17:10 +00:00)


train_class_df.sample(5)

time: 12 ms (started: 2023-03-14 06:17:10 +00:00)


# Let's plot the number of examinations for each class detected, grouped by Target value.
fig, ax = plt.subplots(nrows=1,figsize=(12,6))
tmp = train_class_df.groupby('Target')['class'].value_counts()
df = pd.DataFrame(data={'Exams': tmp.values}, index=tmp.index).reset_index()
sns.barplot(ax=ax,x = 'Target', y='Exams',hue='class',data=df)
plt.title("Chest exams class and Target")
plt.show()

time: 223 ms (started: 2023-03-14 06:17:10 +00:00)


# Density plots for class - Lung Opacity (Target class = 1)
target1 = train_class_df[train_class_df['Target']==1]
sns.set_style('whitegrid')
plt.figure()
fig, ax = plt.subplots(2,2,figsize=(12,12))
sns.distplot(target1['x'],kde=True,bins=50, color="red", ax=ax[0,0])
sns.distplot(target1['y'],kde=True,bins=50, color="blue", ax=ax[0,1])
sns.distplot(target1['width'],kde=True,bins=50, color="green", ax=ax[1,0])
sns.distplot(target1['height'],kde=True,bins=50, color="magenta", ax=ax[1,1])
locs, labels = plt.xticks()
plt.tick_params(axis='both', which='major', labelsize=12)
plt.show()

<Figure size 432x288 with 0 Axes>

time: 1.44 s (started: 2023-03-14 06:17:10 +00:00)


# We will show a sample of center points (xc, yc) superimposed with the corresponding sample of the rectangles.
fig, ax = plt.subplots(1,1,figsize=(7,7))
target_sample = target1.sample(2000)
target_sample['xc'] = target_sample['x'] + target_sample['width'] / 2
target_sample['yc'] = target_sample['y'] + target_sample['height'] / 2
plt.title("Centers of Lung Opacity rectangles (brown) over rectangles (yellow)\nSample size: 2000")
target_sample.plot.scatter(x='xc', y='yc', xlim=(0,1024), ylim=(0,1024), ax=ax, alpha=0.8, marker=".", color="brown")
for i, crt_sample in target_sample.iterrows():
    ax.add_patch(Rectangle(xy=(crt_sample['x'], crt_sample['y']),
                width=crt_sample['width'],height=crt_sample['height'],alpha=3.5e-3, color="yellow"))
plt.show()

time: 4.1 s (started: 2023-03-14 06:17:12 +00:00)


corr = train_class_df.corr()
plt.figure(figsize=(12,5))

sns.heatmap(corr,annot=True)

## There is high corelation between widht and height

<AxesSubplot:>

time: 445 ms (started: 2023-03-14 06:17:16 +00:00)


# The files names are the patients IDs.
image_sample_path = os.listdir('stage_2_train_images')[:5]
print(image_sample_path)

# GDrive I/O Problem; Run after

['f79e5a75-5e15-431a-9ba3-de64c1724b65.dcm', 'f79fa94e-8886-425c-95d4-81abfea92aee.dcm', 'f79fc759-fde0-4e2e-8951-26ac8af5ef89.dcm', 'f7a169e2-a1a7-439a-ab83-0d34cf5ae4b9.dcm', 'f7a172fc-a34f-4155-806d-7b0afe2a6156.dcm']
time: 661 ms (started: 2023-03-13 15:08:47 +00:00)


image_train_path = os.listdir('stage_2_train_images')
image_test_path = os.listdir('stage_2_test_images')
print("Number of images in train set:", len(image_train_path),"\nNumber of images in test set:", len(image_test_path))

Number of images in train set: 26684 
Number of images in test set: 3000
time: 372 ms (started: 2023-03-13 15:08:48 +00:00)


print("Unique patientId in  train_class_df: ", train_class_df['patientId'].nunique())

Unique patientId in  train_class_df:  26684
time: 15.3 ms (started: 2023-03-13 15:08:48 +00:00)


# Duplicate patients in the dataset
duplicateRowsDF = train_class_df[train_class_df.duplicated(['patientId'])]
duplicateRowsDF.shape

(10945, 7)

time: 17.9 ms (started: 2023-03-13 15:08:48 +00:00)


duplicateRowsDF.head(2)

time: 11.1 ms (started: 2023-03-13 15:08:48 +00:00)


## Examining one of the patient id which is duplicate , we can see that the x,y, widht and height is not the same
## This indicates that the same patient has two bounding boxes in the same dicom image
train_class_df[train_class_df.patientId=='00436515-870c-4b36-a041-de91049b9ab4']

time: 18 ms (started: 2023-03-13 15:08:48 +00:00)


train_class_df[train_class_df.patientId=='00704310-78a8-4b38-8475-49f4573b2dbb']

time: 18.6 ms (started: 2023-03-13 15:08:49 +00:00)


# Understand the target, class and duplicate entries
tmp = train_class_df.groupby(['patientId','Target', 'class'])['patientId'].count()
df = pd.DataFrame(data={'Exams': tmp.values}, index=tmp.index).reset_index()
tmp = df.groupby(['Exams','Target','class']).count()
df2 = pd.DataFrame(data=tmp.values, index=tmp.index).reset_index()
df2.columns = ['Exams', 'Target','Class', 'Entries']
df2

time: 69.9 ms (started: 2023-03-13 15:08:49 +00:00)


fig, ax = plt.subplots(nrows=1,figsize=(12,6))
sns.barplot(ax=ax,x = 'Target', y='Entries', hue='Exams',data=df2)
plt.title("Chest exams class and Target")
plt.show()

time: 252 ms (started: 2023-03-13 15:08:49 +00:00)


samplePatientID = list(train_class_df[:3].T.to_dict().values())[0]['patientId']
samplePatientID = samplePatientID+'.dcm'
dicom_file_path = os.path.join("stage_2_train_images/",samplePatientID)
dicom_file_dataset = dcm.read_file(dicom_file_path)
dicom_file_dataset

Dataset.file_meta -------------------------------
(0002, 0000) File Meta Information Group Length  UL: 202
(0002, 0001) File Meta Information Version       OB: b'\x00\x01'
(0002, 0002) Media Storage SOP Class UID         UI: Secondary Capture Image Storage
(0002, 0003) Media Storage SOP Instance UID      UI: 1.2.276.0.7230010.3.1.4.8323329.28530.1517874485.775526
(0002, 0010) Transfer Syntax UID                 UI: JPEG Baseline (Process 1)
(0002, 0012) Implementation Class UID            UI: 1.2.276.0.7230010.3.0.3.6.0
(0002, 0013) Implementation Version Name         SH: 'OFFIS_DCMTK_360'
-------------------------------------------------
(0008, 0005) Specific Character Set              CS: 'ISO_IR 100'
(0008, 0016) SOP Class UID                       UI: Secondary Capture Image Storage
(0008, 0018) SOP Instance UID                    UI: 1.2.276.0.7230010.3.1.4.8323329.28530.1517874485.775526
(0008, 0020) Study Date                          DA: '19010101'
(0008, 0030) Study Time                          TM: '000000.00'
(0008, 0050) Accession Number                    SH: ''
(0008, 0060) Modality                            CS: 'CR'
(0008, 0064) Conversion Type                     CS: 'WSD'
(0008, 0090) Referring Physician's Name          PN: ''
(0008, 103e) Series Description                  LO: 'view: PA'
(0010, 0010) Patient's Name                      PN: '0004cfab-14fd-4e49-80ba-63a80b6bddd6'
(0010, 0020) Patient ID                          LO: '0004cfab-14fd-4e49-80ba-63a80b6bddd6'
(0010, 0030) Patient's Birth Date                DA: ''
(0010, 0040) Patient's Sex                       CS: 'F'
(0010, 1010) Patient's Age                       AS: '51'
(0018, 0015) Body Part Examined                  CS: 'CHEST'
(0018, 5101) View Position                       CS: 'PA'
(0020, 000d) Study Instance UID                  UI: 1.2.276.0.7230010.3.1.2.8323329.28530.1517874485.775525
(0020, 000e) Series Instance UID                 UI: 1.2.276.0.7230010.3.1.3.8323329.28530.1517874485.775524
(0020, 0010) Study ID                            SH: ''
(0020, 0011) Series Number                       IS: '1'
(0020, 0013) Instance Number                     IS: '1'
(0020, 0020) Patient Orientation                 CS: ''
(0028, 0002) Samples per Pixel                   US: 1
(0028, 0004) Photometric Interpretation          CS: 'MONOCHROME2'
(0028, 0010) Rows                                US: 1024
(0028, 0011) Columns                             US: 1024
(0028, 0030) Pixel Spacing                       DS: [0.14300000000000002, 0.14300000000000002]
(0028, 0100) Bits Allocated                      US: 8
(0028, 0101) Bits Stored                         US: 8
(0028, 0102) High Bit                            US: 7
(0028, 0103) Pixel Representation                US: 0
(0028, 2110) Lossy Image Compression             CS: '01'
(0028, 2114) Lossy Image Compression Method      CS: 'ISO_10918_1'
(7fe0, 0010) Pixel Data                          OB: Array of 142006 elements

time: 777 ms (started: 2023-03-10 11:57:22 +00:00)


# Plot DICOM images with Target = 1
def show_dicom_images(data):
    img_data = list(data.T.to_dict().values())
    f, ax = plt.subplots(3,3, figsize=(16,18))
    for i,data_row in enumerate(img_data):
        patientImage = data_row['patientId']+'.dcm'
        imagePath = os.path.join("stage_2_train_images/",patientImage)
        data_row_img_data = dcm.read_file(imagePath)
        modality = data_row_img_data.Modality
        age = data_row_img_data.PatientAge
        sex = data_row_img_data.PatientSex
        data_row_img = dcm.dcmread(imagePath)
        ax[i//3, i%3].imshow(data_row_img.pixel_array, cmap=plt.cm.bone) 
        ax[i//3, i%3].axis('off')
        ax[i//3, i%3].set_title('ID: {}\nModality: {} Age: {} Sex: {} Target: {}\nClass: {}\nWindow: {}:{}:{}:{}'.format(
                data_row['patientId'],
                modality, age, sex, data_row['Target'], data_row['class'], 
                data_row['x'],data_row['y'],data_row['width'],data_row['height']))
    plt.show()

time: 3.65 ms (started: 2023-03-10 11:57:23 +00:00)


show_dicom_images(train_class_df[train_class_df['Target']==1].sample(9))

time: 8.74 s (started: 2023-03-10 11:57:23 +00:00)


# Images with bounding boxes
def show_dicom_images_with_boxes(data):
    img_data = list(data.T.to_dict().values())
    f, ax = plt.subplots(3,3, figsize=(16,18))
    for i,data_row in enumerate(img_data):
        patientImage = data_row['patientId']+'.dcm'
        imagePath = os.path.join("stage_2_train_images/",patientImage)
        data_row_img_data = dcm.read_file(imagePath)
        modality = data_row_img_data.Modality
        age = data_row_img_data.PatientAge
        sex = data_row_img_data.PatientSex
        data_row_img = dcm.dcmread(imagePath)
        ax[i//3, i%3].imshow(data_row_img.pixel_array, cmap=plt.cm.bone) 
        ax[i//3, i%3].axis('off')
        ax[i//3, i%3].set_title('ID: {}\nModality: {} Age: {} Sex: {} Target: {}\nClass: {}'.format(
                data_row['patientId'],modality, age, sex, data_row['Target'], data_row['class']))
        rows = train_class_df[train_class_df['patientId']==data_row['patientId']]
        box_data = list(rows.T.to_dict().values())
        for j, row in enumerate(box_data):
            ax[i//3, i%3].add_patch(Rectangle(xy=(row['x'], row['y']),
                        width=row['width'],height=row['height'], 
                        color="yellow",alpha = 0.1))   
    plt.show()

time: 8.74 ms (started: 2023-03-10 11:57:32 +00:00)


show_dicom_images_with_boxes(train_class_df[train_class_df['Target']==1].sample(9))

time: 7.93 s (started: 2023-03-10 11:57:32 +00:00)


# Plot DICOM images with Target = 0
show_dicom_images(train_class_df[train_class_df['Target']==0].sample(9))

time: 7.98 s (started: 2023-03-10 11:57:40 +00:00)


vars = ['Modality', 'PatientAge', 'PatientSex', 'BodyPartExamined', 'ViewPosition', 'ConversionType', 'Rows', 'Columns', 'PixelSpacing']

def process_dicom_data(data_df, data_path):
    for var in vars:
        data_df[var] = None
    image_names = os.listdir(data_path)
    for i, img_name in tqdm_notebook(enumerate(image_names)):
        imagePath = os.path.join(data_path,img_name)
        data_row_img_data = dcm.read_file(imagePath)
        idx = (data_df['patientId']==data_row_img_data.PatientID)
        data_df.loc[idx,'Modality'] = data_row_img_data.Modality
        data_df.loc[idx,'PatientAge'] = pd.to_numeric(data_row_img_data.PatientAge)
        data_df.loc[idx,'PatientSex'] = data_row_img_data.PatientSex
        data_df.loc[idx,'BodyPartExamined'] = data_row_img_data.BodyPartExamined
        data_df.loc[idx,'ViewPosition'] = data_row_img_data.ViewPosition
        data_df.loc[idx,'ConversionType'] = data_row_img_data.ConversionType
        data_df.loc[idx,'Rows'] = data_row_img_data.Rows
        data_df.loc[idx,'Columns'] = data_row_img_data.Columns  
        data_df.loc[idx,'PixelSpacing'] = str.format("{:4.3f}",data_row_img_data.PixelSpacing[0])

time: 8.67 ms (started: 2023-03-10 11:57:48 +00:00)


process_dicom_data(train_class_df,'stage_2_train_images/')

# max time = 24 min; Run the notebook again if it takes more time

0it [00:00, ?it/s]

time: 24min 3s (started: 2023-03-10 12:30:37 +00:00)


train_class_df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 37629 entries, 0 to 37628
Data columns (total 16 columns):
 #   Column            Non-Null Count  Dtype  
---  ------            --------------  -----  
 0   patientId         37629 non-null  object 
 1   x                 16957 non-null  float64
 2   y                 16957 non-null  float64
 3   width             16957 non-null  float64
 4   height            16957 non-null  float64
 5   Target            37629 non-null  int64  
 6   class             37629 non-null  object 
 7   Modality          37629 non-null  object 
 8   PatientAge        37629 non-null  object 
 9   PatientSex        37629 non-null  object 
 10  BodyPartExamined  37629 non-null  object 
 11  ViewPosition      37629 non-null  object 
 12  ConversionType    37629 non-null  object 
 13  Rows              37629 non-null  object 
 14  Columns           37629 non-null  object 
 15  PixelSpacing      37629 non-null  object 
dtypes: float64(4), int64(1), object(11)
memory usage: 4.9+ MB
time: 71.5 ms (started: 2023-03-10 12:54:41 +00:00)


train_class_df.sample(10)

time: 49 ms (started: 2023-03-10 12:54:41 +00:00)


test_class_df = pd.read_csv('stage_2_sample_submission.csv')

time: 691 ms (started: 2023-03-10 12:54:41 +00:00)


test_class_df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3000 entries, 0 to 2999
Data columns (total 2 columns):
 #   Column            Non-Null Count  Dtype 
---  ------            --------------  ----- 
 0   patientId         3000 non-null   object
 1   PredictionString  3000 non-null   object
dtypes: object(2)
memory usage: 47.0+ KB
time: 12.5 ms (started: 2023-03-10 12:54:41 +00:00)


test_class_df.sample(10)

time: 15.6 ms (started: 2023-03-10 12:54:41 +00:00)


test_class_df = test_class_df.drop('PredictionString',1)
process_dicom_data(test_class_df,'stage_2_test_images/')

0it [00:00, ?it/s]

time: 2min 29s (started: 2023-03-10 12:54:41 +00:00)


test_class_df.sample(10)

time: 17.7 ms (started: 2023-03-10 12:57:10 +00:00)


print("Modalities: train:",train_class_df['Modality'].unique(), "test:", test_class_df['Modality'].unique())

Modalities: train: ['CR'] test: ['CR']
time: 16 ms (started: 2023-03-10 12:57:10 +00:00)


print("Body Part Examined: train:",train_class_df['BodyPartExamined'].unique(), "test:", test_class_df['BodyPartExamined'].unique())

Body Part Examined: train: ['CHEST'] test: ['CHEST']
time: 14.2 ms (started: 2023-03-10 12:57:11 +00:00)


print("Body Part Examined (train):", train_class_df['BodyPartExamined'].value_counts())
print()
print("Body Part Examined (test):", test_class_df['BodyPartExamined'].value_counts())

Body Part Examined (train): CHEST    37629
Name: BodyPartExamined, dtype: int64

Body Part Examined (test): CHEST    3000
Name: BodyPartExamined, dtype: int64
time: 99.4 ms (started: 2023-03-10 14:05:37 +00:00)


print("View Position: train:",train_class_df['ViewPosition'].unique(), "test:", test_class_df['ViewPosition'].unique())

View Position: train: ['PA' 'AP'] test: ['PA' 'AP']
time: 13.7 ms (started: 2023-03-10 12:57:11 +00:00)


f,axes=plt.subplots(1,2,figsize=(17,7))
train_class_df['ViewPosition'].value_counts().plot.pie(autopct='%1.1f%%',ax=axes[0])
sns.countplot('ViewPosition',data=train_class_df,ax=axes[1])
axes[0].set_title('Pie Chart for ViewPosition')
axes[1].set_title('Bar Graph for ViewPosition')
plt.show()

time: 335 ms (started: 2023-03-10 12:57:11 +00:00)


def plot_window(data, color_point, color_window, text):
    fig, ax = plt.subplots(1,1,figsize=(7,7))
    plt.title("Centers of Lung Opacity rectangles over rectangles\n{}".format(text))
    data.plot.scatter(x='xc', y='yc', xlim=(0,1024), ylim=(0,1024), ax=ax, alpha=0.8, marker=".", color=color_point)
    for i, crt_sample in data.iterrows():
        ax.add_patch(Rectangle(xy=(crt_sample['x'], crt_sample['y']),
            width=crt_sample['width'],height=crt_sample['height'],alpha=3.5e-3, color=color_window))
    plt.show()

time: 1.69 ms (started: 2023-03-10 12:57:11 +00:00)


target1 = train_class_df[train_class_df['Target']==1]

target_sample = target1.sample(2000)
target_sample['xc'] = target_sample['x'] + target_sample['width'] / 2
target_sample['yc'] = target_sample['y'] + target_sample['height'] / 2

target_ap = target_sample[target_sample['ViewPosition']=='AP']
target_pa = target_sample[target_sample['ViewPosition']=='PA']

time: 21.9 ms (started: 2023-03-10 12:57:11 +00:00)


plot_window(target_ap,'green', 'yellow', 'Patient View Position: AP')

time: 4 s (started: 2023-03-10 12:57:11 +00:00)


plot_window(target_pa,'blue', 'red', 'Patient View Position: PA')

time: 797 ms (started: 2023-03-10 12:57:15 +00:00)


f,axes=plt.subplots(1,2,figsize=(17,7))
test_class_df['ViewPosition'].value_counts().plot.pie(autopct='%1.1f%%',ax=axes[0])
sns.countplot('ViewPosition',data=test_class_df,ax=axes[1])
axes[0].set_title('Pie Chart for ViewPosition')
axes[1].set_title('Bar Graph for ViewPosition')
plt.show()

time: 284 ms (started: 2023-03-10 12:57:16 +00:00)


print("Conversion Type: train:",train_class_df['ConversionType'].unique(), "test:", test_class_df['ConversionType'].unique())

Conversion Type: train: ['WSD'] test: ['WSD']
time: 14.2 ms (started: 2023-03-10 12:57:16 +00:00)


print("Conversion Type (train):", train_class_df['ConversionType'].value_counts())
print()
print("Conversion Type (test):", test_class_df['ConversionType'].value_counts())

Conversion Type (train): WSD    37629
Name: ConversionType, dtype: int64

Conversion Type (test): WSD    3000
Name: ConversionType, dtype: int64
time: 36.9 ms (started: 2023-03-10 14:08:04 +00:00)


print("Rows: train:",train_class_df['Rows'].unique(), "test:", test_class_df['Rows'].unique())
print("Columns: train:",train_class_df['Columns'].unique(), "test:", test_class_df['Columns'].unique())

Rows: train: [1024] test: [1024]
Columns: train: [1024] test: [1024]
time: 13.4 ms (started: 2023-03-10 12:57:16 +00:00)


print("Rows: train:", train_class_df['Rows'].value_counts())
print()
print("Columns: train:", test_class_df['Rows'].value_counts())

Rows: train: 1024    37629
Name: Rows, dtype: int64

Columns: train: 1024    3000
Name: Rows, dtype: int64
time: 20.8 ms (started: 2023-03-10 14:14:50 +00:00)


print("Rows: train:", train_class_df['Columns'].value_counts())
print()
print("Columns: train:", test_class_df['Columns'].value_counts())

Rows: train: 1024    37629
Name: Columns, dtype: int64

Columns: train: 1024    3000
Name: Columns, dtype: int64
time: 20.6 ms (started: 2023-03-10 14:14:55 +00:00)


tmp = train_class_df.groupby(['Target', 'PatientAge'])['patientId'].count()
df = pd.DataFrame(data={'Exams': tmp.values}, index=tmp.index).reset_index()
tmp = df.groupby(['Exams','Target', 'PatientAge']).count()
df2 = pd.DataFrame(data=tmp.values, index=tmp.index).reset_index()

time: 25.2 ms (started: 2023-03-10 12:57:16 +00:00)


tmp = train_class_df.groupby(['class', 'PatientAge'])['patientId'].count()
df1 = pd.DataFrame(data={'Exams': tmp.values}, index=tmp.index).reset_index()
tmp = df1.groupby(['Exams','class', 'PatientAge']).count()
df3 = pd.DataFrame(data=tmp.values, index=tmp.index).reset_index()

time: 24 ms (started: 2023-03-10 12:57:16 +00:00)


fig, (ax) = plt.subplots(nrows=1,figsize=(16,6))
sns.barplot(ax=ax, x = 'PatientAge', y='Exams', hue='Target',data=df2)
plt.title("Train set: Chest exams Age and Target")
plt.xticks(rotation=90)
plt.show()

time: 2.53 s (started: 2023-03-10 12:57:16 +00:00)


fig, (ax) = plt.subplots(nrows=1,figsize=(16,6))
sns.barplot(ax=ax, x = 'PatientAge', y='Exams', hue='class',data=df3)
plt.title("Train set: Chest exams Age and class")
plt.xticks(rotation=90)
plt.show()

time: 5.8 s (started: 2023-03-10 12:57:19 +00:00)


target_age1 = target_sample[target_sample['PatientAge'] < 20]
target_age2 = target_sample[(target_sample['PatientAge'] >=20) & (target_sample['PatientAge'] < 35)]
target_age3 = target_sample[(target_sample['PatientAge'] >=35) & (target_sample['PatientAge'] < 50)]
target_age4 = target_sample[(target_sample['PatientAge'] >=50) & (target_sample['PatientAge'] < 65)]
target_age5 = target_sample[target_sample['PatientAge'] >= 65]

time: 16.8 ms (started: 2023-03-10 12:57:25 +00:00)


plot_window(target_age1,'blue', 'red', 'Patient Age: 1-19 years')

time: 568 ms (started: 2023-03-10 12:57:25 +00:00)


plot_window(target_age2,'blue', 'red', 'Patient Age: 20-34 years')

time: 1.05 s (started: 2023-03-10 12:57:25 +00:00)


plot_window(target_age3,'blue', 'red', 'Patient Age: 35-49 years')

time: 1.58 s (started: 2023-03-10 12:57:26 +00:00)


plot_window(target_age4,'blue', 'red', 'Patient Age: 50-65 years')

time: 1.33 s (started: 2023-03-10 12:57:28 +00:00)


plot_window(target_age5,'blue', 'red', 'Patient Age: 65+ years')

time: 727 ms (started: 2023-03-10 12:57:29 +00:00)


fig, (ax) = plt.subplots(nrows=1,figsize=(16,6))
sns.countplot(test_class_df['PatientAge'], ax=ax)
plt.title("Test set: Patient Age")
plt.xticks(rotation=90)
plt.show()

time: 1.69 s (started: 2023-03-10 12:57:30 +00:00)


tmp = train_class_df.groupby(['Target', 'PatientSex'])['patientId'].count()
df = pd.DataFrame(data={'Exams': tmp.values}, index=tmp.index).reset_index()
tmp = df.groupby(['Exams','Target', 'PatientSex']).count()
df2 = pd.DataFrame(data=tmp.values, index=tmp.index).reset_index()
fig, ax = plt.subplots(nrows=1,figsize=(6,6))
sns.barplot(ax=ax, x = 'PatientSex', y='Exams', hue='Target',data=df2)
plt.title("Train set: Patient Sex and Target")
plt.show()

time: 211 ms (started: 2023-03-10 12:57:32 +00:00)


tmp = train_class_df.groupby(['class', 'PatientSex'])['patientId'].count()
df1 = pd.DataFrame(data={'Exams': tmp.values}, index=tmp.index).reset_index()
tmp = df1.groupby(['Exams','class', 'PatientSex']).count()
df3 = pd.DataFrame(data=tmp.values, index=tmp.index).reset_index()
fig, (ax) = plt.subplots(nrows=1,figsize=(6,6))
sns.barplot(ax=ax, x = 'PatientSex', y='Exams', hue='class',data=df3)
plt.title("Train set: Patient Sex and class")
plt.show()

time: 230 ms (started: 2023-03-10 12:57:32 +00:00)


target_female = target_sample[target_sample['PatientSex']=='F']
target_male = target_sample[target_sample['PatientSex']=='M']

time: 6.63 ms (started: 2023-03-10 12:57:32 +00:00)


plot_window(target_female,"red", "magenta","Patients Sex: Female")

time: 2.18 s (started: 2023-03-10 12:57:32 +00:00)


plot_window(target_male,"darkblue", "blue", "Patients Sex: Male")

time: 3.58 s (started: 2023-03-10 12:57:34 +00:00)


f,axes=plt.subplots(1,2,figsize=(17,7))
test_class_df['PatientSex'].value_counts().plot.pie(autopct='%1.1f%%',ax=axes[0])
sns.countplot('PatientSex',data=test_class_df,ax=axes[1])
axes[0].set_title('Pie Chart for PatientSex')
axes[1].set_title('Bar Graph for PatientSex')
plt.show()

time: 420 ms (started: 2023-03-10 12:57:38 +00:00)


train_class_df['PixelSpacing'].value_counts()

0.168    14581
0.143     9711
0.139     8813
0.171     3083
0.194     1424
0.199        9
0.115        8
Name: PixelSpacing, dtype: int64

time: 72.7 ms (started: 2023-03-10 16:10:51 +00:00)


sns.countplot(x='PixelSpacing',hue='Target',data=train_class_df)

<AxesSubplot:xlabel='PixelSpacing', ylabel='count'>

time: 359 ms (started: 2023-03-10 16:12:01 +00:00)


sns.countplot(x='PixelSpacing',hue='class',data=train_class_df)

<AxesSubplot:xlabel='PixelSpacing', ylabel='count'>

time: 1.95 s (started: 2023-03-10 16:11:48 +00:00)


test_class_df['PixelSpacing'].value_counts()

0.168    1014
0.143     986
0.139     607
0.171     225
0.194     166
0.115       2
Name: PixelSpacing, dtype: int64

time: 11.8 ms (started: 2023-03-10 16:13:20 +00:00)


sns.countplot(x='PixelSpacing',hue='PatientSex',data=test_class_df)

<AxesSubplot:xlabel='PixelSpacing', ylabel='count'>

time: 925 ms (started: 2023-03-10 16:14:56 +00:00)


sns.countplot(x='PixelSpacing',hue='PatientSex',data=test_class_df)

<AxesSubplot:xlabel='PixelSpacing', ylabel='count'>

time: 497 ms (started: 2023-03-10 16:15:27 +00:00)


# File paths from the G Drive
trainImagesPath = "stage_2_train_images"
testImagesPath = "stage_2_test_images"

labelsPath = "stage_2_train_labels.csv"
classInfoPath = "stage_2_detailed_class_info.csv"

# Read the labels and classinfo
labels = pd.read_csv(labelsPath)
details = pd.read_csv(classInfoPath)

time: 3.07 s (started: 2023-03-25 01:27:04 +00:00)


labels.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 30227 entries, 0 to 30226
Data columns (total 6 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   patientId  30227 non-null  object 
 1   x          9555 non-null   float64
 2   y          9555 non-null   float64
 3   width      9555 non-null   float64
 4   height     9555 non-null   float64
 5   Target     30227 non-null  int64  
dtypes: float64(4), int64(1), object(1)
memory usage: 1.4+ MB
time: 23.8 ms (started: 2023-03-25 01:27:07 +00:00)


labels.sample(10)

time: 17.1 ms (started: 2023-03-25 01:27:07 +00:00)


details.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 30227 entries, 0 to 30226
Data columns (total 2 columns):
 #   Column     Non-Null Count  Dtype 
---  ------     --------------  ----- 
 0   patientId  30227 non-null  object
 1   class      30227 non-null  object
dtypes: object(2)
memory usage: 472.4+ KB
time: 13.8 ms (started: 2023-03-25 01:27:07 +00:00)


details.sample(10)

time: 6.9 ms (started: 2023-03-25 01:27:07 +00:00)


# Concatenate the two datasets - 'labels' and 'details':
train_data = pd.concat([labels, details['class']], axis = 1)

train_data.head()

time: 10.6 ms (started: 2023-03-25 01:27:07 +00:00)


train_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 30227 entries, 0 to 30226
Data columns (total 7 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   patientId  30227 non-null  object 
 1   x          9555 non-null   float64
 2   y          9555 non-null   float64
 3   width      9555 non-null   float64
 4   height     9555 non-null   float64
 5   Target     30227 non-null  int64  
 6   class      30227 non-null  object 
dtypes: float64(4), int64(1), object(2)
memory usage: 1.6+ MB
time: 20.7 ms (started: 2023-03-25 01:27:07 +00:00)


# Using 1000 samples from each class
# Increase the dataset size later
train_data1 = train_data.groupby('class', group_keys=False).apply(lambda x: x.sample(1000))

time: 10.6 ms (started: 2023-03-25 01:27:07 +00:00)


# Check the training dataset with class distribution (data balancing)
train_data1["class"].value_counts()

Lung Opacity                    1000
No Lung Opacity / Not Normal    1000
Normal                          1000
Name: class, dtype: int64

time: 4.31 ms (started: 2023-03-25 01:27:07 +00:00)


# Final train data info
train_data1.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 3000 entries, 29479 to 2923
Data columns (total 7 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   patientId  3000 non-null   object 
 1   x          1000 non-null   float64
 2   y          1000 non-null   float64
 3   width      1000 non-null   float64
 4   height     1000 non-null   float64
 5   Target     3000 non-null   int64  
 6   class      3000 non-null   object 
dtypes: float64(4), int64(1), object(2)
memory usage: 187.5+ KB
time: 9.5 ms (started: 2023-03-25 01:27:07 +00:00)


# Create the arrays
images = []
ADJUSTED_IMAGE_SIZE = 128 # Other image sizes to try later: [(224, 224), (384, 384), (512, 512), (640, 640)]
imageList = []
classLabels = []       
labels = []
originalImage = []

# Function to read the image from the path and reshape the image to smaller size
def readAndReshapeImage(image):
    img = np.array(image).astype(np.uint8)
    # Resize the image
    res = cv2.resize(img,(ADJUSTED_IMAGE_SIZE,ADJUSTED_IMAGE_SIZE), interpolation = cv2.INTER_LINEAR)
    return res

# Read the image and resize the image
def populateImage(rowData):
    for index, row in rowData.iterrows():
        patientId = row.patientId
        classlabel = row["class"]
        dcm_file = 'stage_2_train_images/'+'{}.dcm'.format(patientId)
        dcm_data = dcm.read_file(dcm_file)
        img = dcm_data.pixel_array
        # Convert the images to 3 channels as the dicom image pixels do not have colour class
        if len(img.shape) != 3 or img.shape[2] != 3:
            img = np.stack((img,) * 3, -1)
        imageList.append(readAndReshapeImage(img))
        # originalImage.append(img)
        classLabels.append(classlabel)
    tmpImages = np.array(imageList)
    tmpLabels = np.array(classLabels)
        # originalImages = np.array(originalImage)
    return tmpImages,tmpLabels

time: 3.33 ms (started: 2023-03-25 01:27:07 +00:00)


%%time
## Read the images into numpy arrays
images, labels = populateImage(train_data1)

CPU times: user 34 s, sys: 8.89 s, total: 42.9 s
Wall time: 5min 13s
time: 5min 13s (started: 2023-03-25 01:59:00 +00:00)


# Sample images
images[:1]

array([[[[ 16,  16,  16],
         [  4,   4,   4],
         [  4,   4,   4],
         ...,
         [  4,   4,   4],
         [  4,   4,   4],
         [  5,   5,   5]],

        [[ 15,  15,  15],
         [  4,   4,   4],
         [  4,   4,   4],
         ...,
         [  4,   4,   4],
         [  4,   4,   4],
         [  6,   6,   6]],

        [[ 14,  14,  14],
         [  4,   4,   4],
         [  4,   4,   4],
         ...,
         [  4,   4,   4],
         [  4,   4,   4],
         [  6,   6,   6]],

        ...,

        [[  4,   4,   4],
         [  4,   4,   4],
         [  4,   4,   4],
         ...,
         [  0,   0,   0],
         [ 17,  17,  17],
         [121, 121, 121]],

        [[  5,   5,   5],
         [  5,   5,   5],
         [  4,   4,   4],
         ...,
         [  0,   0,   0],
         [ 18,  18,  18],
         [126, 126, 126]],

        [[ 14,  14,  14],
         [ 11,  11,  11],
         [ 10,  10,  10],
         ...,
         [  2,   2,   2],
         [ 22,  22,  22],
         [134, 134, 134]]]], dtype=uint8)

time: 7.23 ms (started: 2023-03-25 02:04:13 +00:00)


# Sample labels
labels[:10]

array(['Lung Opacity', 'Lung Opacity', 'Lung Opacity', 'Lung Opacity',
       'Lung Opacity', 'Lung Opacity', 'Lung Opacity', 'Lung Opacity',
       'Lung Opacity', 'Lung Opacity'], dtype='<U28')

time: 4.49 ms (started: 2023-03-25 02:04:13 +00:00)


# Shape of the dataset (X, yc)
# The image is of 128*128 with 3 channels
images.shape , labels.shape

((5609, 128, 128, 3), (5609,))

time: 3.17 ms (started: 2023-03-25 02:04:13 +00:00)


# Check the converted image 
plt.imshow(images[12])

<matplotlib.image.AxesImage at 0x7f9493ec9c40>

time: 332 ms (started: 2023-03-25 02:04:13 +00:00)


# Check the unique labels
np.unique(labels), len(np.unique(labels))

(array(['Lung Opacity', 'No Lung Opacity / Not Normal', 'Normal'],
       dtype='<U28'), 3)

time: 10.9 ms (started: 2023-03-25 02:04:14 +00:00)


# Encode the labels
from sklearn.preprocessing import LabelBinarizer
enc = LabelBinarizer()
yc = enc.fit_transform(labels)

time: 18.4 ms (started: 2023-03-25 02:04:14 +00:00)


yc[:10]

array([[1, 0, 0],
       [1, 0, 0],
       [1, 0, 0],
       [1, 0, 0],
       [1, 0, 0],
       [1, 0, 0],
       [1, 0, 0],
       [1, 0, 0],
       [1, 0, 0],
       [1, 0, 0]])

time: 3.74 ms (started: 2023-03-25 02:04:14 +00:00)


# Split the data for training and testing
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(images, yc, test_size=0.3, random_state=0)

time: 97.8 ms (started: 2023-03-25 02:04:14 +00:00)


# Function to plot the various metrics across the epochs
"""
@Description: This function plots our metrics for our models across epochs
@Inputs: The history of the fitted model
@Output: Plots for accuracy, precision, recall, AUC, and loss
"""
def plottingScores(hist):
    fig, ax = plt.subplots(1, 5, figsize=(25, 3))
    ax = ax.ravel()

    for i, met in enumerate(['accuracy', 'precision', 'recall', 'AUC', 'loss']):
        ax[i].plot(hist.history[met])
        ax[i].plot(hist.history['val_' + met])
        ax[i].set_title('Model {}'.format(met))
        ax[i].set_xlabel('epochs')
        ax[i].set_ylabel(met)
        ax[i].legend(['train', 'val'])

time: 2.54 ms (started: 2023-03-25 02:04:14 +00:00)


# Metrics to evaluate the model
METRICS = ['accuracy', 
           tf.keras.metrics.Precision(name='precision'), 
           tf.keras.metrics.Recall(name='recall'), 
           tf.keras.metrics.AUC(name='AUC')]

time: 2.86 s (started: 2023-03-25 02:04:14 +00:00)


# Define the callback, checkpoint and early_stopping
def exponential_decay(lr0, s):
    def exponential_decay_fn(epoch):
        return lr0 * 0.1 **(epoch / s)
    return exponential_decay_fn

exponential_decay_fn = exponential_decay(0.01, 20)
lr_scheduler = tf.keras.callbacks.LearningRateScheduler(exponential_decay_fn)
checkpoint_cb = tf.keras.callbacks.ModelCheckpoint("cr_model.h5", save_best_only=True)
early_stopping_cb = tf.keras.callbacks.EarlyStopping(patience=5, restore_best_weights=True)

time: 12.5 ms (started: 2023-03-25 02:04:17 +00:00)


# Function to create a simple fully connected NN
def fcnn_model():
    # Basic model with a flattening layer followed by 2 dense layers
    model = tf.keras.models.Sequential([
                tf.keras.layers.Flatten(input_shape = (128, 128, 3)), 
                tf.keras.layers.Dense(128, activation = "relu"), 
                tf.keras.layers.Dense(3, activation = "softmax")
                ])
    
    return model

time: 4.65 ms (started: 2023-03-25 02:04:17 +00:00)


# Build our FCNN model and compile
model1 = fcnn_model()
model1.summary()
model1.compile(optimizer="adam", loss="categorical_crossentropy", metrics=METRICS)

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 flatten (Flatten)           (None, 49152)             0         
                                                                 
 dense (Dense)               (None, 128)               6291584   
                                                                 
 dense_1 (Dense)             (None, 3)                 387       
                                                                 
=================================================================
Total params: 6,291,971
Trainable params: 6,291,971
Non-trainable params: 0
_________________________________________________________________
time: 164 ms (started: 2023-03-25 02:04:17 +00:00)


# Fit the model
H = model1.fit(X_train, y_train,  
                          epochs = 30,
                          batch_size = 128,
                          validation_split = 0.2, 
                          # class_weight = classWeight, 
                          verbose = 1,
                          callbacks = [checkpoint_cb, early_stopping_cb, lr_scheduler])

Epoch 1/30
25/25 [==============================] - 8s 136ms/step - loss: 49849.2070 - accuracy: 0.3459 - precision: 0.3459 - recall: 0.3459 - AUC: 0.5094 - val_loss: 2102.8447 - val_accuracy: 0.5433 - val_precision: 0.5433 - val_recall: 0.5433 - val_AUC: 0.6574 - lr: 0.0100
Epoch 2/30
25/25 [==============================] - 1s 22ms/step - loss: 1260.3652 - accuracy: 0.4815 - precision: 0.4815 - recall: 0.4815 - AUC: 0.6108 - val_loss: 504.2290 - val_accuracy: 0.5216 - val_precision: 0.5216 - val_recall: 0.5216 - val_AUC: 0.6404 - lr: 0.0089
Epoch 3/30
25/25 [==============================] - 1s 22ms/step - loss: 331.9548 - accuracy: 0.5414 - precision: 0.5414 - recall: 0.5414 - AUC: 0.6567 - val_loss: 263.9551 - val_accuracy: 0.4847 - val_precision: 0.4847 - val_recall: 0.4847 - val_AUC: 0.6135 - lr: 0.0079
Epoch 4/30
25/25 [==============================] - 1s 22ms/step - loss: 166.4397 - accuracy: 0.5535 - precision: 0.5535 - recall: 0.5535 - AUC: 0.6658 - val_loss: 177.8865 - val_accuracy: 0.4491 - val_precision: 0.4491 - val_recall: 0.4491 - val_AUC: 0.5885 - lr: 0.0071
Epoch 5/30
25/25 [==============================] - 1s 22ms/step - loss: 128.2540 - accuracy: 0.5354 - precision: 0.5354 - recall: 0.5354 - AUC: 0.6535 - val_loss: 125.8664 - val_accuracy: 0.5522 - val_precision: 0.5522 - val_recall: 0.5522 - val_AUC: 0.6657 - lr: 0.0063
Epoch 6/30
25/25 [==============================] - 0s 12ms/step - loss: 142.2826 - accuracy: 0.5006 - precision: 0.5006 - recall: 0.5006 - AUC: 0.6267 - val_loss: 186.9144 - val_accuracy: 0.5089 - val_precision: 0.5089 - val_recall: 0.5089 - val_AUC: 0.6346 - lr: 0.0056
Epoch 7/30
25/25 [==============================] - 0s 12ms/step - loss: 166.7835 - accuracy: 0.5029 - precision: 0.5029 - recall: 0.5029 - AUC: 0.6282 - val_loss: 136.7881 - val_accuracy: 0.5204 - val_precision: 0.5204 - val_recall: 0.5204 - val_AUC: 0.6414 - lr: 0.0050
Epoch 8/30
25/25 [==============================] - 1s 21ms/step - loss: 177.2953 - accuracy: 0.4682 - precision: 0.4682 - recall: 0.4682 - AUC: 0.6027 - val_loss: 125.7408 - val_accuracy: 0.5165 - val_precision: 0.5165 - val_recall: 0.5165 - val_AUC: 0.6434 - lr: 0.0045
Epoch 9/30
25/25 [==============================] - 1s 23ms/step - loss: 102.6262 - accuracy: 0.5557 - precision: 0.5557 - recall: 0.5557 - AUC: 0.6702 - val_loss: 62.5748 - val_accuracy: 0.6005 - val_precision: 0.6005 - val_recall: 0.6005 - val_AUC: 0.7012 - lr: 0.0040
Epoch 10/30
25/25 [==============================] - 1s 32ms/step - loss: 59.0157 - accuracy: 0.5790 - precision: 0.5790 - recall: 0.5790 - AUC: 0.6882 - val_loss: 44.1414 - val_accuracy: 0.5852 - val_precision: 0.5852 - val_recall: 0.5852 - val_AUC: 0.6911 - lr: 0.0035
Epoch 11/30
25/25 [==============================] - 0s 18ms/step - loss: 44.1616 - accuracy: 0.5847 - precision: 0.5847 - recall: 0.5847 - AUC: 0.6946 - val_loss: 133.1523 - val_accuracy: 0.4237 - val_precision: 0.4237 - val_recall: 0.4237 - val_AUC: 0.5675 - lr: 0.0032
Epoch 12/30
25/25 [==============================] - 1s 39ms/step - loss: 48.2316 - accuracy: 0.5685 - precision: 0.5685 - recall: 0.5685 - AUC: 0.6789 - val_loss: 32.1840 - val_accuracy: 0.6069 - val_precision: 0.6069 - val_recall: 0.6069 - val_AUC: 0.7077 - lr: 0.0028
Epoch 13/30
25/25 [==============================] - 0s 15ms/step - loss: 31.2740 - accuracy: 0.6150 - precision: 0.6150 - recall: 0.6150 - AUC: 0.7161 - val_loss: 52.5623 - val_accuracy: 0.4618 - val_precision: 0.4618 - val_recall: 0.4618 - val_AUC: 0.6010 - lr: 0.0025
Epoch 14/30
25/25 [==============================] - 0s 15ms/step - loss: 39.3380 - accuracy: 0.5764 - precision: 0.5764 - recall: 0.5764 - AUC: 0.6871 - val_loss: 46.6306 - val_accuracy: 0.4720 - val_precision: 0.4720 - val_recall: 0.4720 - val_AUC: 0.6071 - lr: 0.0022
Epoch 15/30
25/25 [==============================] - 0s 14ms/step - loss: 24.5999 - accuracy: 0.6213 - precision: 0.6213 - recall: 0.6213 - AUC: 0.7268 - val_loss: 33.3471 - val_accuracy: 0.4962 - val_precision: 0.4962 - val_recall: 0.4962 - val_AUC: 0.6303 - lr: 0.0020
Epoch 16/30
25/25 [==============================] - 0s 15ms/step - loss: 27.7948 - accuracy: 0.5936 - precision: 0.5936 - recall: 0.5936 - AUC: 0.7020 - val_loss: 38.3336 - val_accuracy: 0.5420 - val_precision: 0.5420 - val_recall: 0.5420 - val_AUC: 0.6650 - lr: 0.0018
Epoch 17/30
25/25 [==============================] - 1s 21ms/step - loss: 24.2190 - accuracy: 0.6185 - precision: 0.6185 - recall: 0.6185 - AUC: 0.7250 - val_loss: 28.1696 - val_accuracy: 0.5802 - val_precision: 0.5802 - val_recall: 0.5802 - val_AUC: 0.6927 - lr: 0.0016
Epoch 18/30
25/25 [==============================] - 0s 12ms/step - loss: 21.3103 - accuracy: 0.6239 - precision: 0.6239 - recall: 0.6239 - AUC: 0.7277 - val_loss: 38.2093 - val_accuracy: 0.5242 - val_precision: 0.5242 - val_recall: 0.5242 - val_AUC: 0.6473 - lr: 0.0014
Epoch 19/30
25/25 [==============================] - 1s 23ms/step - loss: 20.3377 - accuracy: 0.6210 - precision: 0.6212 - recall: 0.6210 - AUC: 0.7275 - val_loss: 19.3079 - val_accuracy: 0.6107 - val_precision: 0.6102 - val_recall: 0.6094 - val_AUC: 0.7190 - lr: 0.0013
Epoch 20/30
25/25 [==============================] - 1s 21ms/step - loss: 11.3019 - accuracy: 0.6930 - precision: 0.6932 - recall: 0.6930 - AUC: 0.7874 - val_loss: 19.2675 - val_accuracy: 0.5941 - val_precision: 0.5941 - val_recall: 0.5941 - val_AUC: 0.7140 - lr: 0.0011
Epoch 21/30
25/25 [==============================] - 0s 12ms/step - loss: 12.8136 - accuracy: 0.6605 - precision: 0.6605 - recall: 0.6605 - AUC: 0.7609 - val_loss: 30.3743 - val_accuracy: 0.5000 - val_precision: 0.5000 - val_recall: 0.5000 - val_AUC: 0.6322 - lr: 0.0010
Epoch 22/30
25/25 [==============================] - 1s 21ms/step - loss: 12.9175 - accuracy: 0.6717 - precision: 0.6717 - recall: 0.6717 - AUC: 0.7679 - val_loss: 14.9079 - val_accuracy: 0.6183 - val_precision: 0.6183 - val_recall: 0.6183 - val_AUC: 0.7261 - lr: 8.9125e-04
Epoch 23/30
25/25 [==============================] - 0s 11ms/step - loss: 8.4172 - accuracy: 0.7166 - precision: 0.7166 - recall: 0.7166 - AUC: 0.8095 - val_loss: 16.8035 - val_accuracy: 0.5483 - val_precision: 0.5483 - val_recall: 0.5483 - val_AUC: 0.6765 - lr: 7.9433e-04
Epoch 24/30
25/25 [==============================] - 1s 21ms/step - loss: 7.8858 - accuracy: 0.7239 - precision: 0.7241 - recall: 0.7239 - AUC: 0.8088 - val_loss: 13.8908 - val_accuracy: 0.6323 - val_precision: 0.6323 - val_recall: 0.6323 - val_AUC: 0.7368 - lr: 7.0795e-04
Epoch 25/30
25/25 [==============================] - 1s 23ms/step - loss: 8.9721 - accuracy: 0.6936 - precision: 0.6936 - recall: 0.6936 - AUC: 0.7880 - val_loss: 13.6134 - val_accuracy: 0.6323 - val_precision: 0.6323 - val_recall: 0.6323 - val_AUC: 0.7378 - lr: 6.3096e-04
Epoch 26/30
25/25 [==============================] - 0s 13ms/step - loss: 6.5074 - accuracy: 0.7439 - precision: 0.7439 - recall: 0.7439 - AUC: 0.8275 - val_loss: 14.9198 - val_accuracy: 0.5802 - val_precision: 0.5802 - val_recall: 0.5802 - val_AUC: 0.6953 - lr: 5.6234e-04
Epoch 27/30
25/25 [==============================] - 0s 11ms/step - loss: 6.1100 - accuracy: 0.7494 - precision: 0.7493 - recall: 0.7490 - AUC: 0.8350 - val_loss: 13.7376 - val_accuracy: 0.5751 - val_precision: 0.5751 - val_recall: 0.5751 - val_AUC: 0.6915 - lr: 5.0119e-04
Epoch 28/30
25/25 [==============================] - 1s 22ms/step - loss: 7.8541 - accuracy: 0.7000 - precision: 0.7000 - recall: 0.7000 - AUC: 0.7936 - val_loss: 11.8359 - val_accuracy: 0.6196 - val_precision: 0.6204 - val_recall: 0.6196 - val_AUC: 0.7337 - lr: 4.4668e-04
Epoch 29/30
25/25 [==============================] - 0s 12ms/step - loss: 6.0399 - accuracy: 0.7369 - precision: 0.7372 - recall: 0.7369 - AUC: 0.8256 - val_loss: 14.4477 - val_accuracy: 0.5802 - val_precision: 0.5802 - val_recall: 0.5802 - val_AUC: 0.7060 - lr: 3.9811e-04
Epoch 30/30
25/25 [==============================] - 0s 12ms/step - loss: 5.4317 - accuracy: 0.7554 - precision: 0.7557 - recall: 0.7554 - AUC: 0.8404 - val_loss: 12.6684 - val_accuracy: 0.6158 - val_precision: 0.6158 - val_recall: 0.6158 - val_AUC: 0.7301 - lr: 3.5481e-04
time: 22.8 s (started: 2023-03-25 02:04:17 +00:00)


# Evaluate and display results
results = model1.evaluate(X_test, y_test) # Evaluate the model on test data
results = dict(zip(model1.metrics_names,results))

print(results)
plottingScores(H)

53/53 [==============================] - 0s 4ms/step - loss: 10.9302 - accuracy: 0.6328 - precision: 0.6328 - recall: 0.6328 - AUC: 0.7454
{'loss': 10.930243492126465, 'accuracy': 0.6327985525131226, 'precision': 0.6327985525131226, 'recall': 0.6327985525131226, 'AUC': 0.7454316020011902}


y_pred = model1.predict(X_test)
# y_pred = np.array([0 if proba < 0.5 else 1 for proba in y_pred_proba])

time: 1.95 s (started: 2023-03-25 02:04:40 +00:00)
53/53 [==============================] - 0s 2ms/step
time: 442 ms (started: 2023-03-25 02:04:42 +00:00)


y_pred[:10]

array([[9.1158631e-10, 9.9880219e-01, 1.1978684e-03],
       [1.0000000e+00, 1.9673133e-10, 1.3054741e-17],
       [4.2965649e-25, 1.0000000e+00, 9.6033059e-16],
       [1.6006921e-30, 1.0000000e+00, 7.9788942e-33],
       [1.7161180e-17, 1.0000000e+00, 4.4823514e-17],
       [9.9985075e-01, 1.4928424e-04, 1.9262859e-22],
       [2.3723891e-01, 4.0507304e-09, 7.6276112e-01],
       [9.9999225e-01, 7.7055684e-06, 0.0000000e+00],
       [3.5582500e-19, 9.9513948e-01, 4.8604743e-03],
       [9.9999559e-01, 4.3661971e-06, 0.0000000e+00]], dtype=float32)

time: 4.46 ms (started: 2023-03-25 02:04:42 +00:00)


# Classification Report
print("Classification Report:\n",classification_report(y_test.argmax(axis=1), y_pred.argmax(axis=1)))

# Confusion Matrix
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test.argmax(axis=1), y_pred.argmax(axis=1))
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1', '2']],  
                         columns = [i for i in ['0', '1', '2']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.67      0.75      0.71       603
           1       0.55      0.62      0.58       600
           2       0.75      0.50      0.60       480

    accuracy                           0.63      1683
   macro avg       0.65      0.62      0.63      1683
weighted avg       0.65      0.63      0.63      1683

Confusion Matrix Chart:

time: 326 ms (started: 2023-03-25 02:04:42 +00:00)


# Model comparison
precision = precision_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
recall = recall_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
f1 = f1_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
AUC = results['AUC']

Train_Accuracy = model1.evaluate(X_train, y_train)
Test_Accuracy = model1.evaluate(X_test, y_test)

base_1 = []
base_1.append(['Model-1: Base Model', Train_Accuracy[1], Test_Accuracy[1], precision, recall, f1, AUC])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

123/123 [==============================] - 1s 4ms/step - loss: 6.7049 - accuracy: 0.7264 - precision: 0.7264 - recall: 0.7264 - AUC: 0.8210
53/53 [==============================] - 0s 5ms/step - loss: 10.9302 - accuracy: 0.6328 - precision: 0.6328 - recall: 0.6328 - AUC: 0.7454
time: 1.33 s (started: 2023-03-25 02:04:43 +00:00)


# Training history dataframe
hist = pd.DataFrame(H.history)

# Understand the various Metrics with epochs
hist

time: 31.6 ms (started: 2023-03-25 02:04:44 +00:00)


hist.describe()

time: 51.8 ms (started: 2023-03-25 02:04:44 +00:00)


# Function to create a complex NN model
def cnn_model1():
    model = tf.keras.Sequential([
        tf.keras.layers.Conv2D(32, kernel_size=(3,3), strides=(1,1), padding = 'valid', activation = 'relu', input_shape=(128, 128, 3)), #  convolutional layer
        tf.keras.layers.MaxPool2D(pool_size=(2,2)), # flatten output of conv
        
        tf.keras.layers.Conv2D(32, kernel_size=(3,3), strides=(1,1), padding = 'valid', activation = 'relu'), #  convolutional layer
        tf.keras.layers.MaxPool2D(pool_size=(2,2)), # flatten output of conv
        tf.keras.layers.Dropout(0.3),
        
        tf.keras.layers.Conv2D(64, 3, activation = 'relu', padding = 'valid'),
        tf.keras.layers.Conv2D(128, 3, activation = 'relu', padding = 'valid'),
        tf.keras.layers.BatchNormalization(),
        tf.keras.layers.MaxPool2D(),
        tf.keras.layers.Dropout(0.4),

        tf.keras.layers.Flatten(), # flatten output of conv
        tf.keras.layers.Dense(512, activation = "relu"), # hidden layer
        tf.keras.layers.BatchNormalization(),
        tf.keras.layers.Dropout(0.5),
        tf.keras.layers.Dense(128, activation = "relu"), #  output layer
        tf.keras.layers.BatchNormalization(),
        tf.keras.layers.Dropout(0.5),
        tf.keras.layers.Dense(3, activation = "softmax")])
    
    return model

time: 6.65 ms (started: 2023-03-25 02:04:44 +00:00)


# Build and compile the model
model2 = cnn_model1()
model2.summary()
model2.compile(optimizer="adam", loss="categorical_crossentropy", metrics=METRICS)

Model: "sequential_1"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 conv2d (Conv2D)             (None, 126, 126, 32)      896       
                                                                 
 max_pooling2d (MaxPooling2D  (None, 63, 63, 32)       0         
 )                                                               
                                                                 
 conv2d_1 (Conv2D)           (None, 61, 61, 32)        9248      
                                                                 
 max_pooling2d_1 (MaxPooling  (None, 30, 30, 32)       0         
 2D)                                                             
                                                                 
 dropout (Dropout)           (None, 30, 30, 32)        0         
                                                                 
 conv2d_2 (Conv2D)           (None, 28, 28, 64)        18496     
                                                                 
 conv2d_3 (Conv2D)           (None, 26, 26, 128)       73856     
                                                                 
 batch_normalization (BatchN  (None, 26, 26, 128)      512       
 ormalization)                                                   
                                                                 
 max_pooling2d_2 (MaxPooling  (None, 13, 13, 128)      0         
 2D)                                                             
                                                                 
 dropout_1 (Dropout)         (None, 13, 13, 128)       0         
                                                                 
 flatten_1 (Flatten)         (None, 21632)             0         
                                                                 
 dense_2 (Dense)             (None, 512)               11076096  
                                                                 
 batch_normalization_1 (Batc  (None, 512)              2048      
 hNormalization)                                                 
                                                                 
 dropout_2 (Dropout)         (None, 512)               0         
                                                                 
 dense_3 (Dense)             (None, 128)               65664     
                                                                 
 batch_normalization_2 (Batc  (None, 128)              512       
 hNormalization)                                                 
                                                                 
 dropout_3 (Dropout)         (None, 128)               0         
                                                                 
 dense_4 (Dense)             (None, 3)                 387       
                                                                 
=================================================================
Total params: 11,247,715
Trainable params: 11,246,179
Non-trainable params: 1,536
_________________________________________________________________
time: 319 ms (started: 2023-03-25 02:04:44 +00:00)


# Fit the model
H = model2.fit(X_train, y_train,  
                      epochs=30, 
                      validation_split = 0.20, 
                      batch_size=128,
                      # class_weight=classWeight,
                      callbacks=[checkpoint_cb, early_stopping_cb, lr_scheduler],
                      verbose=1)

Epoch 1/30
25/25 [==============================] - 14s 125ms/step - loss: 1.2800 - accuracy: 0.4701 - precision: 0.5456 - recall: 0.4717 - AUC: 0.7001 - val_loss: 56.0337 - val_accuracy: 0.3372 - val_precision: 0.3372 - val_recall: 0.3372 - val_AUC: 0.5047 - lr: 0.0100
Epoch 2/30
25/25 [==============================] - 2s 83ms/step - loss: 0.9630 - accuracy: 0.5449 - precision: 0.5818 - recall: 0.4204 - AUC: 0.7352 - val_loss: 16.2147 - val_accuracy: 0.3193 - val_precision: 0.3193 - val_recall: 0.3193 - val_AUC: 0.4869 - lr: 0.0089
Epoch 3/30
25/25 [==============================] - 4s 151ms/step - loss: 0.9170 - accuracy: 0.5516 - precision: 0.5970 - recall: 0.4146 - AUC: 0.7524 - val_loss: 4.3532 - val_accuracy: 0.3422 - val_precision: 0.3436 - val_recall: 0.3422 - val_AUC: 0.5323 - lr: 0.0079
Epoch 4/30
25/25 [==============================] - 3s 103ms/step - loss: 0.8938 - accuracy: 0.5666 - precision: 0.6187 - recall: 0.4341 - AUC: 0.7647 - val_loss: 1.3821 - val_accuracy: 0.4135 - val_precision: 0.4410 - val_recall: 0.3855 - val_AUC: 0.6490 - lr: 0.0071
Epoch 5/30
25/25 [==============================] - 3s 104ms/step - loss: 0.8992 - accuracy: 0.5583 - precision: 0.6038 - recall: 0.4382 - AUC: 0.7619 - val_loss: 1.3046 - val_accuracy: 0.4402 - val_precision: 0.4466 - val_recall: 0.4313 - val_AUC: 0.6477 - lr: 0.0063
Epoch 6/30
25/25 [==============================] - 3s 103ms/step - loss: 0.8789 - accuracy: 0.5723 - precision: 0.6204 - recall: 0.4331 - AUC: 0.7694 - val_loss: 1.2886 - val_accuracy: 0.4669 - val_precision: 0.4862 - val_recall: 0.3804 - val_AUC: 0.6508 - lr: 0.0056
Epoch 7/30
25/25 [==============================] - 3s 113ms/step - loss: 0.8629 - accuracy: 0.5841 - precision: 0.6337 - recall: 0.4475 - AUC: 0.7790 - val_loss: 0.9547 - val_accuracy: 0.5522 - val_precision: 0.5773 - val_recall: 0.4847 - val_AUC: 0.7449 - lr: 0.0050
Epoch 8/30
25/25 [==============================] - 3s 105ms/step - loss: 0.8289 - accuracy: 0.6029 - precision: 0.6525 - recall: 0.4844 - AUC: 0.7983 - val_loss: 0.9314 - val_accuracy: 0.5382 - val_precision: 0.5733 - val_recall: 0.4427 - val_AUC: 0.7439 - lr: 0.0045
Epoch 9/30
25/25 [==============================] - 3s 103ms/step - loss: 0.8422 - accuracy: 0.5850 - precision: 0.6362 - recall: 0.4662 - AUC: 0.7889 - val_loss: 0.8887 - val_accuracy: 0.5712 - val_precision: 0.6271 - val_recall: 0.4644 - val_AUC: 0.7713 - lr: 0.0040
Epoch 10/30
25/25 [==============================] - 3s 102ms/step - loss: 0.8277 - accuracy: 0.6159 - precision: 0.6566 - recall: 0.5035 - AUC: 0.8006 - val_loss: 0.8629 - val_accuracy: 0.6005 - val_precision: 0.6420 - val_recall: 0.4517 - val_AUC: 0.7851 - lr: 0.0035
Epoch 11/30
25/25 [==============================] - 3s 103ms/step - loss: 0.8226 - accuracy: 0.6041 - precision: 0.6460 - recall: 0.4987 - AUC: 0.8013 - val_loss: 0.8518 - val_accuracy: 0.5954 - val_precision: 0.6655 - val_recall: 0.4707 - val_AUC: 0.7908 - lr: 0.0032
Epoch 12/30
25/25 [==============================] - 3s 113ms/step - loss: 0.8119 - accuracy: 0.6083 - precision: 0.6644 - recall: 0.4911 - AUC: 0.8069 - val_loss: 0.8481 - val_accuracy: 0.6056 - val_precision: 0.6672 - val_recall: 0.4975 - val_AUC: 0.7948 - lr: 0.0028
Epoch 13/30
25/25 [==============================] - 3s 110ms/step - loss: 0.8075 - accuracy: 0.6242 - precision: 0.6690 - recall: 0.5232 - AUC: 0.8096 - val_loss: 0.8388 - val_accuracy: 0.6183 - val_precision: 0.6634 - val_recall: 0.5140 - val_AUC: 0.7995 - lr: 0.0025
Epoch 14/30
25/25 [==============================] - 2s 77ms/step - loss: 0.8043 - accuracy: 0.6309 - precision: 0.6714 - recall: 0.5172 - AUC: 0.8137 - val_loss: 0.8617 - val_accuracy: 0.5967 - val_precision: 0.6617 - val_recall: 0.5127 - val_AUC: 0.7905 - lr: 0.0022
Epoch 15/30
25/25 [==============================] - 2s 79ms/step - loss: 0.8027 - accuracy: 0.6169 - precision: 0.6617 - recall: 0.5220 - AUC: 0.8128 - val_loss: 0.8458 - val_accuracy: 0.6031 - val_precision: 0.6594 - val_recall: 0.5000 - val_AUC: 0.7959 - lr: 0.0020
Epoch 16/30
25/25 [==============================] - 3s 104ms/step - loss: 0.7873 - accuracy: 0.6315 - precision: 0.6888 - recall: 0.5315 - AUC: 0.8213 - val_loss: 0.8227 - val_accuracy: 0.6272 - val_precision: 0.6794 - val_recall: 0.5204 - val_AUC: 0.8081 - lr: 0.0018
Epoch 17/30
25/25 [==============================] - 2s 79ms/step - loss: 0.7760 - accuracy: 0.6360 - precision: 0.6808 - recall: 0.5462 - AUC: 0.8265 - val_loss: 0.8429 - val_accuracy: 0.5992 - val_precision: 0.6586 - val_recall: 0.5204 - val_AUC: 0.7992 - lr: 0.0016
Epoch 18/30
25/25 [==============================] - 2s 81ms/step - loss: 0.7830 - accuracy: 0.6357 - precision: 0.6780 - recall: 0.5484 - AUC: 0.8236 - val_loss: 0.8275 - val_accuracy: 0.6310 - val_precision: 0.6689 - val_recall: 0.5140 - val_AUC: 0.8102 - lr: 0.0014
Epoch 19/30
25/25 [==============================] - 2s 83ms/step - loss: 0.7621 - accuracy: 0.6465 - precision: 0.6944 - recall: 0.5557 - AUC: 0.8321 - val_loss: 0.8332 - val_accuracy: 0.6323 - val_precision: 0.6677 - val_recall: 0.5267 - val_AUC: 0.8091 - lr: 0.0013
Epoch 20/30
25/25 [==============================] - 2s 80ms/step - loss: 0.7712 - accuracy: 0.6439 - precision: 0.6837 - recall: 0.5541 - AUC: 0.8282 - val_loss: 0.8420 - val_accuracy: 0.6056 - val_precision: 0.6537 - val_recall: 0.5356 - val_AUC: 0.8049 - lr: 0.0011
Epoch 21/30
25/25 [==============================] - 2s 82ms/step - loss: 0.7709 - accuracy: 0.6433 - precision: 0.6769 - recall: 0.5484 - AUC: 0.8278 - val_loss: 0.8478 - val_accuracy: 0.5878 - val_precision: 0.6422 - val_recall: 0.5115 - val_AUC: 0.7977 - lr: 0.0010
time: 1min 2s (started: 2023-03-25 02:04:44 +00:00)


# Evaluate the model results and put into a dict
results = model2.evaluate(X_test, y_test)
results = dict(zip(model2.metrics_names,results))

print(results)
plottingScores(H)

53/53 [==============================] - 1s 9ms/step - loss: 0.7753 - accuracy: 0.6471 - precision: 0.6931 - recall: 0.5288 - AUC: 0.8272
{'loss': 0.7753071188926697, 'accuracy': 0.6470588445663452, 'precision': 0.6931464076042175, 'recall': 0.5288175940513611, 'AUC': 0.8272116780281067}

time: 24.7 s (started: 2023-03-25 02:05:47 +00:00)


y_pred = model2.predict(X_test)
# y_pred = np.array([0 if proba < 0.5 else 1 for proba in y_pred_proba])

53/53 [==============================] - 0s 5ms/step
time: 600 ms (started: 2023-03-25 02:06:12 +00:00)


# Classification Report
print("Classification Report:\n",classification_report(y_test.argmax(axis=1), y_pred.argmax(axis=1)))

# Confusion Matrix
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test.argmax(axis=1), y_pred.argmax(axis=1))
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1', '2']],  
                         columns = [i for i in ['0', '1', '2']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.65      0.81      0.72       603
           1       0.56      0.43      0.49       600
           2       0.73      0.70      0.72       480

    accuracy                           0.65      1683
   macro avg       0.65      0.65      0.64      1683
weighted avg       0.64      0.65      0.64      1683

Confusion Matrix Chart:

time: 301 ms (started: 2023-03-25 02:06:12 +00:00)


# Model comparison
precision = precision_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
recall = recall_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
f1 = f1_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
AUC = results['AUC']

Train_Accuracy = model2.evaluate(X_train, y_train)
Test_Accuracy = model2.evaluate(X_test, y_test)

# base_1 = []
base_1.append(['Model-2: CNN-1', Train_Accuracy[1], Test_Accuracy[1], precision, recall, f1, AUC])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

123/123 [==============================] - 1s 9ms/step - loss: 0.7561 - accuracy: 0.6538 - precision: 0.7082 - recall: 0.5367 - AUC: 0.8376
53/53 [==============================] - 0s 9ms/step - loss: 0.7753 - accuracy: 0.6471 - precision: 0.6931 - recall: 0.5288 - AUC: 0.8272
time: 2.39 s (started: 2023-03-25 02:06:13 +00:00)


# Function to create a complex NN model
def cnn_model2():
    model = tf.keras.Sequential([
        tf.keras.layers.Conv2D(25, kernel_size=(3,3), strides=(1,1), padding='valid', activation='relu', input_shape=(128, 128, 3)), #  convolutional layer
        tf.keras.layers.MaxPool2D(pool_size=(2,2)), # flatten output of conv
        
        tf.keras.layers.Conv2D(25, kernel_size=(3,3), strides=(1,1), padding='valid', activation='relu'), #  convolutional layer
        tf.keras.layers.BatchNormalization(),
        tf.keras.layers.MaxPool2D(pool_size=(2,2)), # flatten output of conv
        tf.keras.layers.Dropout(0.2),
        
        tf.keras.layers.Conv2D(32, 3, activation='relu', padding='valid'),
        tf.keras.layers.Conv2D(64, 3, activation='relu', padding='valid'),
        tf.keras.layers.BatchNormalization(),
        tf.keras.layers.MaxPool2D(),
        tf.keras.layers.Dropout(0.3),
        

        tf.keras.layers.Flatten(), # Flattening
        
        # Full Connection
        tf.keras.layers.Dense(64, activation='relu'), # hidden layer
        tf.keras.layers.Dropout(0.5), # Dropout
        tf.keras.layers.Dense(3, activation='softmax')]) #  output layer
    
    return model

time: 3.39 ms (started: 2023-03-25 02:06:15 +00:00)


# Build and compile the model
model3 = cnn_model2()
model3.summary()
model3.compile(optimizer="adam", loss="categorical_crossentropy", metrics=METRICS)

Model: "sequential_2"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 conv2d_4 (Conv2D)           (None, 126, 126, 25)      700       
                                                                 
 max_pooling2d_3 (MaxPooling  (None, 63, 63, 25)       0         
 2D)                                                             
                                                                 
 conv2d_5 (Conv2D)           (None, 61, 61, 25)        5650      
                                                                 
 batch_normalization_3 (Batc  (None, 61, 61, 25)       100       
 hNormalization)                                                 
                                                                 
 max_pooling2d_4 (MaxPooling  (None, 30, 30, 25)       0         
 2D)                                                             
                                                                 
 dropout_4 (Dropout)         (None, 30, 30, 25)        0         
                                                                 
 conv2d_6 (Conv2D)           (None, 28, 28, 32)        7232      
                                                                 
 conv2d_7 (Conv2D)           (None, 26, 26, 64)        18496     
                                                                 
 batch_normalization_4 (Batc  (None, 26, 26, 64)       256       
 hNormalization)                                                 
                                                                 
 max_pooling2d_5 (MaxPooling  (None, 13, 13, 64)       0         
 2D)                                                             
                                                                 
 dropout_5 (Dropout)         (None, 13, 13, 64)        0         
                                                                 
 flatten_2 (Flatten)         (None, 10816)             0         
                                                                 
 dense_5 (Dense)             (None, 64)                692288    
                                                                 
 dropout_6 (Dropout)         (None, 64)                0         
                                                                 
 dense_6 (Dense)             (None, 3)                 195       
                                                                 
=================================================================
Total params: 724,917
Trainable params: 724,739
Non-trainable params: 178
_________________________________________________________________
time: 233 ms (started: 2023-03-25 02:06:15 +00:00)


# Fit the model
H = model3.fit(X_train, y_train,  
                      epochs=30, 
                      validation_split = 0.20, 
                      batch_size=128,
                      # class_weight=classWeight,
                      callbacks=[checkpoint_cb, early_stopping_cb, lr_scheduler],
                      verbose=1)

Epoch 1/30
25/25 [==============================] - 7s 97ms/step - loss: 5.1520 - accuracy: 0.3914 - precision: 0.5694 - recall: 0.3239 - AUC: 0.6598 - val_loss: 127.9711 - val_accuracy: 0.3791 - val_precision: 0.3796 - val_recall: 0.3791 - val_AUC: 0.5348 - lr: 0.0100
Epoch 2/30
25/25 [==============================] - 1s 58ms/step - loss: 1.1065 - accuracy: 0.4248 - precision: 0.6238 - recall: 0.1003 - AUC: 0.6039 - val_loss: 37.8964 - val_accuracy: 0.3944 - val_precision: 0.3946 - val_recall: 0.3931 - val_AUC: 0.5464 - lr: 0.0089
Epoch 3/30
25/25 [==============================] - 2s 61ms/step - loss: 1.0431 - accuracy: 0.4427 - precision: 0.6477 - recall: 0.1271 - AUC: 0.6360 - val_loss: 15.2578 - val_accuracy: 0.3690 - val_precision: 0.3702 - val_recall: 0.3664 - val_AUC: 0.5441 - lr: 0.0079
Epoch 4/30
25/25 [==============================] - 1s 58ms/step - loss: 1.0234 - accuracy: 0.4529 - precision: 0.6574 - recall: 0.1576 - AUC: 0.6534 - val_loss: 3.3653 - val_accuracy: 0.3333 - val_precision: 0.3396 - val_recall: 0.1603 - val_AUC: 0.5157 - lr: 0.0071
Epoch 5/30
25/25 [==============================] - 1s 58ms/step - loss: 1.0208 - accuracy: 0.4414 - precision: 0.6836 - recall: 0.1287 - AUC: 0.6429 - val_loss: 1.8499 - val_accuracy: 0.3410 - val_precision: 0.3632 - val_recall: 0.1081 - val_AUC: 0.5606 - lr: 0.0063
Epoch 6/30
25/25 [==============================] - 1s 58ms/step - loss: 0.9884 - accuracy: 0.4666 - precision: 0.6663 - recall: 0.1710 - AUC: 0.6686 - val_loss: 1.1922 - val_accuracy: 0.4313 - val_precision: 0.5505 - val_recall: 0.1387 - val_AUC: 0.6234 - lr: 0.0056
Epoch 7/30
25/25 [==============================] - 1s 60ms/step - loss: 0.9852 - accuracy: 0.4704 - precision: 0.6869 - recall: 0.1768 - AUC: 0.6735 - val_loss: 1.1085 - val_accuracy: 0.4822 - val_precision: 0.6163 - val_recall: 0.2023 - val_AUC: 0.6676 - lr: 0.0050
Epoch 8/30
25/25 [==============================] - 2s 63ms/step - loss: 0.9658 - accuracy: 0.4793 - precision: 0.6835 - recall: 0.1857 - AUC: 0.6829 - val_loss: 0.9946 - val_accuracy: 0.5267 - val_precision: 0.7553 - val_recall: 0.2277 - val_AUC: 0.7309 - lr: 0.0045
Epoch 9/30
25/25 [==============================] - 2s 65ms/step - loss: 0.9512 - accuracy: 0.5035 - precision: 0.6930 - recall: 0.2293 - AUC: 0.7069 - val_loss: 0.9730 - val_accuracy: 0.5356 - val_precision: 0.7134 - val_recall: 0.2913 - val_AUC: 0.7496 - lr: 0.0040
Epoch 10/30
25/25 [==============================] - 2s 64ms/step - loss: 0.9451 - accuracy: 0.5105 - precision: 0.7116 - recall: 0.2271 - AUC: 0.7098 - val_loss: 0.9552 - val_accuracy: 0.5827 - val_precision: 0.7331 - val_recall: 0.3041 - val_AUC: 0.7690 - lr: 0.0035
Epoch 11/30
25/25 [==============================] - 1s 59ms/step - loss: 0.9390 - accuracy: 0.4978 - precision: 0.6688 - recall: 0.2303 - AUC: 0.7148 - val_loss: 0.9858 - val_accuracy: 0.5229 - val_precision: 0.7613 - val_recall: 0.2354 - val_AUC: 0.7346 - lr: 0.0032
Epoch 12/30
25/25 [==============================] - 1s 59ms/step - loss: 0.9260 - accuracy: 0.5188 - precision: 0.7107 - recall: 0.2401 - AUC: 0.7214 - val_loss: 0.9593 - val_accuracy: 0.5662 - val_precision: 0.7794 - val_recall: 0.2697 - val_AUC: 0.7718 - lr: 0.0028
Epoch 13/30
25/25 [==============================] - 2s 62ms/step - loss: 0.9092 - accuracy: 0.5226 - precision: 0.7180 - recall: 0.2481 - AUC: 0.7319 - val_loss: 0.9528 - val_accuracy: 0.5560 - val_precision: 0.7619 - val_recall: 0.3053 - val_AUC: 0.7642 - lr: 0.0025
Epoch 14/30
25/25 [==============================] - 1s 60ms/step - loss: 0.9155 - accuracy: 0.5210 - precision: 0.7178 - recall: 0.2487 - AUC: 0.7302 - val_loss: 0.9355 - val_accuracy: 0.5980 - val_precision: 0.7500 - val_recall: 0.3282 - val_AUC: 0.7829 - lr: 0.0022
Epoch 15/30
25/25 [==============================] - 2s 62ms/step - loss: 0.9110 - accuracy: 0.5185 - precision: 0.7009 - recall: 0.2650 - AUC: 0.7323 - val_loss: 0.9519 - val_accuracy: 0.5738 - val_precision: 0.7570 - val_recall: 0.3092 - val_AUC: 0.7700 - lr: 0.0020
Epoch 16/30
25/25 [==============================] - 2s 62ms/step - loss: 0.9050 - accuracy: 0.5277 - precision: 0.7172 - recall: 0.2592 - AUC: 0.7339 - val_loss: 0.9774 - val_accuracy: 0.5191 - val_precision: 0.6882 - val_recall: 0.2303 - val_AUC: 0.7446 - lr: 0.0018
Epoch 17/30
25/25 [==============================] - 2s 60ms/step - loss: 0.8978 - accuracy: 0.5268 - precision: 0.7025 - recall: 0.2685 - AUC: 0.7392 - val_loss: 0.9428 - val_accuracy: 0.5776 - val_precision: 0.7543 - val_recall: 0.3359 - val_AUC: 0.7783 - lr: 0.0016
Epoch 18/30
25/25 [==============================] - 2s 65ms/step - loss: 0.8986 - accuracy: 0.5287 - precision: 0.7204 - recall: 0.2675 - AUC: 0.7410 - val_loss: 0.9234 - val_accuracy: 0.5967 - val_precision: 0.7317 - val_recall: 0.3817 - val_AUC: 0.7895 - lr: 0.0014
Epoch 19/30
25/25 [==============================] - 2s 65ms/step - loss: 0.8991 - accuracy: 0.5248 - precision: 0.7116 - recall: 0.2656 - AUC: 0.7387 - val_loss: 0.9354 - val_accuracy: 0.5967 - val_precision: 0.7363 - val_recall: 0.3588 - val_AUC: 0.7891 - lr: 0.0013
Epoch 20/30
25/25 [==============================] - 2s 61ms/step - loss: 0.8849 - accuracy: 0.5382 - precision: 0.7144 - recall: 0.2812 - AUC: 0.7490 - val_loss: 0.9386 - val_accuracy: 0.5954 - val_precision: 0.7626 - val_recall: 0.3270 - val_AUC: 0.7880 - lr: 0.0011
Epoch 21/30
25/25 [==============================] - 1s 59ms/step - loss: 0.8781 - accuracy: 0.5385 - precision: 0.7206 - recall: 0.2809 - AUC: 0.7509 - val_loss: 0.9378 - val_accuracy: 0.5929 - val_precision: 0.7458 - val_recall: 0.3397 - val_AUC: 0.7851 - lr: 0.0010
Epoch 22/30
25/25 [==============================] - 2s 62ms/step - loss: 0.8830 - accuracy: 0.5404 - precision: 0.7106 - recall: 0.2799 - AUC: 0.7494 - val_loss: 0.9396 - val_accuracy: 0.5954 - val_precision: 0.7447 - val_recall: 0.3601 - val_AUC: 0.7868 - lr: 8.9125e-04
Epoch 23/30
25/25 [==============================] - 1s 60ms/step - loss: 0.8699 - accuracy: 0.5427 - precision: 0.7146 - recall: 0.2943 - AUC: 0.7568 - val_loss: 0.9270 - val_accuracy: 0.6018 - val_precision: 0.7494 - val_recall: 0.3766 - val_AUC: 0.7949 - lr: 7.9433e-04
time: 41.1 s (started: 2023-03-25 02:06:15 +00:00)


# Evaluate the model results and put into a dict
results = model3.evaluate(X_test, y_test)
results = dict(zip(model3.metrics_names,results))

print(results)
plottingScores(H)

53/53 [==============================] - 1s 8ms/step - loss: 0.8387 - accuracy: 0.6126 - precision: 0.7167 - recall: 0.3832 - AUC: 0.8001
{'loss': 0.8386739492416382, 'accuracy': 0.6125965714454651, 'precision': 0.7166666388511658, 'recall': 0.38324421644210815, 'AUC': 0.8001126050949097}

time: 4.57 s (started: 2023-03-25 02:06:57 +00:00)


y_pred = model3.predict(X_test)
# y_pred = np.array([0 if proba < 0.5 else 1 for proba in y_pred_proba])

53/53 [==============================] - 0s 5ms/step
time: 695 ms (started: 2023-03-25 02:07:01 +00:00)


# Classification Report
print("Classification Report:\n",classification_report(y_test.argmax(axis=1), y_pred.argmax(axis=1)))

# Confusion Matrix
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test.argmax(axis=1), y_pred.argmax(axis=1))
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1', '2']],  
                         columns = [i for i in ['0', '1', '2']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.67      0.76      0.71       603
           1       0.50      0.48      0.49       600
           2       0.68      0.59      0.63       480

    accuracy                           0.61      1683
   macro avg       0.62      0.61      0.61      1683
weighted avg       0.61      0.61      0.61      1683

Confusion Matrix Chart:

time: 496 ms (started: 2023-03-25 02:07:02 +00:00)


# Model comparison
precision = precision_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
recall = recall_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
f1 = f1_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
AUC = results['AUC']

Train_Accuracy = model3.evaluate(X_train, y_train)
Test_Accuracy = model3.evaluate(X_test, y_test)

# base_1 = []
base_1.append(['Model-3: CNN-2', Train_Accuracy[1], Test_Accuracy[1], precision, recall, f1, AUC])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

123/123 [==============================] - 1s 9ms/step - loss: 0.8196 - accuracy: 0.6093 - precision: 0.7391 - recall: 0.3788 - AUC: 0.8089
53/53 [==============================] - 0s 8ms/step - loss: 0.8387 - accuracy: 0.6126 - precision: 0.7167 - recall: 0.3832 - AUC: 0.8001
time: 2.47 s (started: 2023-03-25 02:07:02 +00:00)


# # Pre-process the data for VGG16 model
# X_train = preprocess_input(X_train) 
# X_test = preprocess_input(X_test)

time: 399 µs (started: 2023-03-25 02:07:05 +00:00)


# Function to create NN with transfer learning
def mn_model():
    model = tf.keras.Sequential([
        tf.keras.applications.MobileNetV2(include_top = False, weights="imagenet", input_shape=(128, 128, 3)),
        tf.keras.layers.GlobalAveragePooling2D(),
        Dense(3, activation = 'softmax')])
    
    model.layers[0].trainable = False
    
    return model

time: 1.26 ms (started: 2023-03-25 02:07:05 +00:00)


# Build and compile the mobile net model
model4 = mn_model()
model4.summary()
model4.compile(optimizer='adam', loss='categorical_crossentropy', metrics=METRICS)

Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_128_no_top.h5
9406464/9406464 [==============================] - 0s 0us/step
Model: "sequential_3"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 mobilenetv2_1.00_128 (Funct  (None, 4, 4, 1280)       2257984   
 ional)                                                          
                                                                 
 global_average_pooling2d (G  (None, 1280)             0         
 lobalAveragePooling2D)                                          
                                                                 
 dense_7 (Dense)             (None, 3)                 3843      
                                                                 
=================================================================
Total params: 2,261,827
Trainable params: 3,843
Non-trainable params: 2,257,984
_________________________________________________________________
time: 1.96 s (started: 2023-03-25 02:07:05 +00:00)


H = model4.fit(X_train, y_train,  
                          epochs = 30, 
                          validation_split = 0.20, 
                          # class_weight = classWeight,
                          batch_size = 64,
                          callbacks = [checkpoint_cb, early_stopping_cb, lr_scheduler])

Epoch 1/30
50/50 [==============================] - 7s 70ms/step - loss: 1.3988 - accuracy: 0.4691 - precision: 0.5363 - recall: 0.4074 - AUC: 0.6906 - val_loss: 0.9356 - val_accuracy: 0.5674 - val_precision: 0.6103 - val_recall: 0.4504 - val_AUC: 0.7550 - lr: 0.0100
Epoch 2/30
50/50 [==============================] - 2s 45ms/step - loss: 0.9025 - accuracy: 0.5564 - precision: 0.6055 - recall: 0.4424 - AUC: 0.7598 - val_loss: 0.9130 - val_accuracy: 0.5687 - val_precision: 0.6153 - val_recall: 0.4618 - val_AUC: 0.7614 - lr: 0.0089
Epoch 3/30
50/50 [==============================] - 2s 43ms/step - loss: 0.8513 - accuracy: 0.5901 - precision: 0.6340 - recall: 0.4898 - AUC: 0.7878 - val_loss: 0.9313 - val_accuracy: 0.5216 - val_precision: 0.5468 - val_recall: 0.4682 - val_AUC: 0.7466 - lr: 0.0079
Epoch 4/30
50/50 [==============================] - 2s 41ms/step - loss: 0.7954 - accuracy: 0.6210 - precision: 0.6813 - recall: 0.5140 - AUC: 0.8170 - val_loss: 0.9714 - val_accuracy: 0.5598 - val_precision: 0.6000 - val_recall: 0.4847 - val_AUC: 0.7510 - lr: 0.0071
Epoch 5/30
50/50 [==============================] - 2s 41ms/step - loss: 0.8143 - accuracy: 0.6185 - precision: 0.6611 - recall: 0.5417 - AUC: 0.8103 - val_loss: 1.0303 - val_accuracy: 0.4822 - val_precision: 0.4859 - val_recall: 0.4618 - val_AUC: 0.7255 - lr: 0.0063
Epoch 6/30
50/50 [==============================] - 2s 42ms/step - loss: 0.7907 - accuracy: 0.6217 - precision: 0.6637 - recall: 0.5398 - AUC: 0.8180 - val_loss: 0.8861 - val_accuracy: 0.6005 - val_precision: 0.6571 - val_recall: 0.5216 - val_AUC: 0.7910 - lr: 0.0056
Epoch 7/30
50/50 [==============================] - 2s 42ms/step - loss: 0.7399 - accuracy: 0.6525 - precision: 0.7045 - recall: 0.5640 - AUC: 0.8430 - val_loss: 0.8773 - val_accuracy: 0.6183 - val_precision: 0.6689 - val_recall: 0.5191 - val_AUC: 0.7908 - lr: 0.0050
Epoch 8/30
50/50 [==============================] - 2s 40ms/step - loss: 0.7130 - accuracy: 0.6726 - precision: 0.7345 - recall: 0.5771 - AUC: 0.8574 - val_loss: 0.9033 - val_accuracy: 0.5992 - val_precision: 0.6581 - val_recall: 0.5242 - val_AUC: 0.7893 - lr: 0.0045
Epoch 9/30
50/50 [==============================] - 2s 45ms/step - loss: 0.7144 - accuracy: 0.6790 - precision: 0.7305 - recall: 0.5834 - AUC: 0.8570 - val_loss: 0.8495 - val_accuracy: 0.6336 - val_precision: 0.6979 - val_recall: 0.5407 - val_AUC: 0.8036 - lr: 0.0040
Epoch 10/30
50/50 [==============================] - 2s 44ms/step - loss: 0.6916 - accuracy: 0.6933 - precision: 0.7540 - recall: 0.5914 - AUC: 0.8679 - val_loss: 0.8317 - val_accuracy: 0.6527 - val_precision: 0.6906 - val_recall: 0.5394 - val_AUC: 0.8119 - lr: 0.0035
Epoch 11/30
50/50 [==============================] - 2s 34ms/step - loss: 0.6794 - accuracy: 0.6885 - precision: 0.7428 - recall: 0.6041 - AUC: 0.8717 - val_loss: 0.9092 - val_accuracy: 0.5954 - val_precision: 0.6459 - val_recall: 0.5267 - val_AUC: 0.7864 - lr: 0.0032
Epoch 12/30
50/50 [==============================] - 2s 34ms/step - loss: 0.6998 - accuracy: 0.6742 - precision: 0.7336 - recall: 0.5876 - AUC: 0.8626 - val_loss: 0.8464 - val_accuracy: 0.6387 - val_precision: 0.6770 - val_recall: 0.5573 - val_AUC: 0.8077 - lr: 0.0028
Epoch 13/30
50/50 [==============================] - 2s 35ms/step - loss: 0.6694 - accuracy: 0.6997 - precision: 0.7597 - recall: 0.6083 - AUC: 0.8770 - val_loss: 0.8286 - val_accuracy: 0.6539 - val_precision: 0.7055 - val_recall: 0.5547 - val_AUC: 0.8149 - lr: 0.0025
Epoch 14/30
50/50 [==============================] - 2s 34ms/step - loss: 0.6665 - accuracy: 0.6987 - precision: 0.7583 - recall: 0.6096 - AUC: 0.8791 - val_loss: 0.8434 - val_accuracy: 0.6081 - val_precision: 0.6571 - val_recall: 0.5509 - val_AUC: 0.8057 - lr: 0.0022
Epoch 15/30
50/50 [==============================] - 2s 34ms/step - loss: 0.6517 - accuracy: 0.7089 - precision: 0.7698 - recall: 0.6115 - AUC: 0.8868 - val_loss: 0.8692 - val_accuracy: 0.5929 - val_precision: 0.6314 - val_recall: 0.5318 - val_AUC: 0.7932 - lr: 0.0020
Epoch 16/30
50/50 [==============================] - 2s 42ms/step - loss: 0.6598 - accuracy: 0.7099 - precision: 0.7621 - recall: 0.6111 - AUC: 0.8823 - val_loss: 0.8787 - val_accuracy: 0.6234 - val_precision: 0.6820 - val_recall: 0.5267 - val_AUC: 0.8012 - lr: 0.0018
Epoch 17/30
50/50 [==============================] - 2s 43ms/step - loss: 0.6480 - accuracy: 0.7159 - precision: 0.7776 - recall: 0.6191 - AUC: 0.8887 - val_loss: 0.8340 - val_accuracy: 0.6539 - val_precision: 0.7058 - val_recall: 0.5585 - val_AUC: 0.8157 - lr: 0.0016
Epoch 18/30
50/50 [==============================] - 2s 37ms/step - loss: 0.6458 - accuracy: 0.7105 - precision: 0.7648 - recall: 0.6236 - AUC: 0.8879 - val_loss: 0.8279 - val_accuracy: 0.6590 - val_precision: 0.7110 - val_recall: 0.5573 - val_AUC: 0.8175 - lr: 0.0014
Epoch 19/30
50/50 [==============================] - 2s 41ms/step - loss: 0.6325 - accuracy: 0.7213 - precision: 0.7841 - recall: 0.6293 - AUC: 0.8944 - val_loss: 0.8701 - val_accuracy: 0.5980 - val_precision: 0.6291 - val_recall: 0.5394 - val_AUC: 0.7939 - lr: 0.0013
Epoch 20/30
50/50 [==============================] - 2s 34ms/step - loss: 0.6540 - accuracy: 0.7029 - precision: 0.7539 - recall: 0.6274 - AUC: 0.8826 - val_loss: 0.8357 - val_accuracy: 0.6578 - val_precision: 0.6887 - val_recall: 0.5433 - val_AUC: 0.8148 - lr: 0.0011
Epoch 21/30
50/50 [==============================] - 2s 34ms/step - loss: 0.6306 - accuracy: 0.7283 - precision: 0.7931 - recall: 0.6360 - AUC: 0.8961 - val_loss: 0.8471 - val_accuracy: 0.6107 - val_precision: 0.6486 - val_recall: 0.5496 - val_AUC: 0.8058 - lr: 0.0010
Epoch 22/30
50/50 [==============================] - 2s 34ms/step - loss: 0.6225 - accuracy: 0.7334 - precision: 0.7929 - recall: 0.6328 - AUC: 0.8999 - val_loss: 0.8321 - val_accuracy: 0.6514 - val_precision: 0.6921 - val_recall: 0.5776 - val_AUC: 0.8161 - lr: 8.9125e-04
Epoch 23/30
50/50 [==============================] - 2s 44ms/step - loss: 0.6266 - accuracy: 0.7223 - precision: 0.7869 - recall: 0.6268 - AUC: 0.8973 - val_loss: 0.8436 - val_accuracy: 0.6158 - val_precision: 0.6677 - val_recall: 0.5547 - val_AUC: 0.8086 - lr: 7.9433e-04
time: 50.3 s (started: 2023-03-25 02:07:07 +00:00)


# Show results and print graphs
results = model4.evaluate(X_test, y_test)
results = dict(zip(model4.metrics_names,results))

print(results)
plottingScores(H)

53/53 [==============================] - 2s 24ms/step - loss: 0.7901 - accuracy: 0.6381 - precision: 0.6954 - recall: 0.5520 - AUC: 0.8265
{'loss': 0.7901402711868286, 'accuracy': 0.638146162033081, 'precision': 0.6953592896461487, 'recall': 0.5519905090332031, 'AUC': 0.8264728784561157}

time: 3.55 s (started: 2023-03-25 02:07:57 +00:00)


y_pred = model4.predict(X_test)
# y_pred = np.array([0 if proba < 0.5 else 1 for proba in y_pred_proba])

53/53 [==============================] - 1s 14ms/step
time: 1.64 s (started: 2023-03-25 02:08:01 +00:00)


# Classification Report
print("Classification Report:\n",classification_report(y_test.argmax(axis=1), y_pred.argmax(axis=1)))

# Confusion Matrix
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test.argmax(axis=1), y_pred.argmax(axis=1))
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1', '2']],  
                         columns = [i for i in ['0', '1', '2']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.64      0.81      0.71       603
           1       0.59      0.41      0.49       600
           2       0.67      0.71      0.69       480

    accuracy                           0.64      1683
   macro avg       0.64      0.64      0.63      1683
weighted avg       0.63      0.64      0.63      1683

Confusion Matrix Chart:

time: 584 ms (started: 2023-03-25 02:08:02 +00:00)


# Model comparison
precision = precision_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
recall = recall_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
f1 = f1_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
AUC = results['AUC']

Train_Accuracy = model4.evaluate(X_train, y_train)
Test_Accuracy = model4.evaluate(X_test, y_test)

# base_1 = []
base_1.append(['Model-4: MobileNetV2', Train_Accuracy[1], Test_Accuracy[1], precision, recall, f1, AUC])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

123/123 [==============================] - 2s 18ms/step - loss: 0.6706 - accuracy: 0.7096 - precision: 0.7749 - recall: 0.6060 - AUC: 0.8794
53/53 [==============================] - 1s 16ms/step - loss: 0.7901 - accuracy: 0.6381 - precision: 0.6954 - recall: 0.5520 - AUC: 0.8265
time: 4.07 s (started: 2023-03-25 02:08:03 +00:00)


# Function to create NN with transfer learning
def vgg16_model():
    model = tf.keras.Sequential([
        tf.keras.applications.VGG16(include_top = False, weights="imagenet", input_shape=(128, 128, 3)),
        tf.keras.layers.GlobalAveragePooling2D(),
        Dense(3, activation = 'softmax')])
    
    model.layers[0].trainable = False
    
    return model

time: 1.19 ms (started: 2023-03-25 02:08:07 +00:00)


# Build and compile the model
model5 = vgg16_model()
model5.summary()
model5.compile(optimizer='adam', loss='categorical_crossentropy', metrics=METRICS)

Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5
58889256/58889256 [==============================] - 3s 0us/step
Model: "sequential_4"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 vgg16 (Functional)          (None, 4, 4, 512)         14714688  
                                                                 
 global_average_pooling2d_1   (None, 512)              0         
 (GlobalAveragePooling2D)                                        
                                                                 
 dense_8 (Dense)             (None, 3)                 1539      
                                                                 
=================================================================
Total params: 14,716,227
Trainable params: 1,539
Non-trainable params: 14,714,688
_________________________________________________________________
time: 4.19 s (started: 2023-03-25 02:08:07 +00:00)


H = model5.fit(X_train, y_train,  
                          epochs = 10, 
                          validation_split = 0.20, 
                          # class_weight = classWeight,
                          batch_size = 64,
                          callbacks = [checkpoint_cb, early_stopping_cb, lr_scheduler])
# t = 80min

Epoch 1/10
50/50 [==============================] - 12s 156ms/step - loss: 3.4869 - accuracy: 0.5436 - precision: 0.5900 - recall: 0.5459 - AUC: 0.7216 - val_loss: 1.8338 - val_accuracy: 0.6349 - val_precision: 0.6352 - val_recall: 0.6247 - val_AUC: 0.8009 - lr: 0.0100
Epoch 2/10
50/50 [==============================] - 5s 107ms/step - loss: 1.5981 - accuracy: 0.6274 - precision: 0.6292 - recall: 0.6210 - AUC: 0.7977 - val_loss: 1.4161 - val_accuracy: 0.6565 - val_precision: 0.6584 - val_recall: 0.6450 - val_AUC: 0.8161 - lr: 0.0089
Epoch 3/10
50/50 [==============================] - 5s 100ms/step - loss: 1.1366 - accuracy: 0.6535 - precision: 0.6597 - recall: 0.6471 - AUC: 0.8365 - val_loss: 2.1875 - val_accuracy: 0.5102 - val_precision: 0.5155 - val_recall: 0.5076 - val_AUC: 0.7167 - lr: 0.0079
Epoch 4/10
50/50 [==============================] - 5s 107ms/step - loss: 1.1974 - accuracy: 0.6535 - precision: 0.6579 - recall: 0.6481 - AUC: 0.8350 - val_loss: 1.4233 - val_accuracy: 0.6120 - val_precision: 0.6196 - val_recall: 0.6031 - val_AUC: 0.7851 - lr: 0.0071
Epoch 5/10
50/50 [==============================] - 5s 108ms/step - loss: 0.7597 - accuracy: 0.7156 - precision: 0.7265 - recall: 0.7064 - AUC: 0.8863 - val_loss: 1.1008 - val_accuracy: 0.6552 - val_precision: 0.6667 - val_recall: 0.6489 - val_AUC: 0.8317 - lr: 0.0063
Epoch 6/10
50/50 [==============================] - 5s 102ms/step - loss: 0.6708 - accuracy: 0.7334 - precision: 0.7401 - recall: 0.7210 - AUC: 0.8997 - val_loss: 2.0167 - val_accuracy: 0.5980 - val_precision: 0.6013 - val_recall: 0.5929 - val_AUC: 0.7579 - lr: 0.0056
Epoch 7/10
50/50 [==============================] - 5s 101ms/step - loss: 0.9696 - accuracy: 0.6873 - precision: 0.6925 - recall: 0.6799 - AUC: 0.8613 - val_loss: 1.3217 - val_accuracy: 0.6336 - val_precision: 0.6398 - val_recall: 0.6260 - val_AUC: 0.7968 - lr: 0.0050
Epoch 8/10
50/50 [==============================] - 5s 106ms/step - loss: 0.6822 - accuracy: 0.7366 - precision: 0.7445 - recall: 0.7258 - AUC: 0.9005 - val_loss: 1.1577 - val_accuracy: 0.6539 - val_precision: 0.6636 - val_recall: 0.6425 - val_AUC: 0.8198 - lr: 0.0045
Epoch 9/10
50/50 [==============================] - 5s 100ms/step - loss: 0.6108 - accuracy: 0.7573 - precision: 0.7670 - recall: 0.7475 - AUC: 0.9127 - val_loss: 1.4489 - val_accuracy: 0.5738 - val_precision: 0.5788 - val_recall: 0.5700 - val_AUC: 0.7759 - lr: 0.0040
Epoch 10/10
50/50 [==============================] - 5s 107ms/step - loss: 0.6313 - accuracy: 0.7548 - precision: 0.7645 - recall: 0.7392 - AUC: 0.9074 - val_loss: 1.2286 - val_accuracy: 0.6310 - val_precision: 0.6387 - val_recall: 0.6209 - val_AUC: 0.8217 - lr: 0.0035
time: 58.8 s (started: 2023-03-25 02:08:11 +00:00)


# Show results and print graphs
results = model5.evaluate(X_test, y_test)
results = dict(zip(model5.metrics_names,results))

print(results)
plottingScores(H)

53/53 [==============================] - 5s 61ms/step - loss: 1.0582 - accuracy: 0.6447 - precision: 0.6538 - recall: 0.6352 - AUC: 0.8298
{'loss': 1.0582020282745361, 'accuracy': 0.6446821093559265, 'precision': 0.6538226008415222, 'recall': 0.6351752877235413, 'AUC': 0.8298393487930298}

time: 31.5 s (started: 2023-03-25 02:09:10 +00:00)


y_pred = model5.predict(X_test)
# y_pred = np.array([0 if proba < 0.5 else 1 for proba in y_pred_proba])

53/53 [==============================] - 2s 40ms/step
time: 2.84 s (started: 2023-03-25 02:09:42 +00:00)


# Classification Report
print("Classification Report:\n",classification_report(y_test.argmax(axis=1), y_pred.argmax(axis=1)))

# Confusion Matrix
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test.argmax(axis=1), y_pred.argmax(axis=1))
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1', '2']],  
                         columns = [i for i in ['0', '1', '2']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.67      0.70      0.68       603
           1       0.56      0.47      0.51       600
           2       0.69      0.80      0.74       480

    accuracy                           0.64      1683
   macro avg       0.64      0.65      0.64      1683
weighted avg       0.64      0.64      0.64      1683

Confusion Matrix Chart:

time: 461 ms (started: 2023-03-25 02:09:44 +00:00)


# Model comparison
precision = precision_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
recall = recall_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
f1 = f1_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
AUC = results['AUC']

Train_Accuracy = model5.evaluate(X_train, y_train)
Test_Accuracy = model5.evaluate(X_test, y_test)

# base_1 = []
base_1.append(['Model-5: VGG16', Train_Accuracy[1], Test_Accuracy[1], precision, recall, f1, AUC])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

123/123 [==============================] - 6s 51ms/step - loss: 0.7104 - accuracy: 0.7392 - precision: 0.7482 - recall: 0.7275 - AUC: 0.8962
53/53 [==============================] - 2s 41ms/step - loss: 1.0582 - accuracy: 0.6447 - precision: 0.6538 - recall: 0.6352 - AUC: 0.8298
time: 9.12 s (started: 2023-03-25 02:09:45 +00:00)


# Function to create NN with transfer learning
def resnet50_model():
    model = tf.keras.Sequential([
        tf.keras.applications.ResNet50(include_top = False, weights="imagenet", input_shape=(128, 128, 3)),
        tf.keras.layers.GlobalAveragePooling2D(),
        Dense(3, activation = 'softmax')])
    
    model.layers[0].trainable = False
    
    return model

time: 984 µs (started: 2023-03-25 02:09:54 +00:00)


# Build and compile the model
model6 = resnet50_model()
model6.summary()
model6.compile(optimizer='adam', loss='categorical_crossentropy', metrics=METRICS)

Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5
94765736/94765736 [==============================] - 4s 0us/step
Model: "sequential_5"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 resnet50 (Functional)       (None, 4, 4, 2048)        23587712  
                                                                 
 global_average_pooling2d_2   (None, 2048)             0         
 (GlobalAveragePooling2D)                                        
                                                                 
 dense_9 (Dense)             (None, 3)                 6147      
                                                                 
=================================================================
Total params: 23,593,859
Trainable params: 6,147
Non-trainable params: 23,587,712
_________________________________________________________________
time: 6.72 s (started: 2023-03-25 02:09:54 +00:00)


H = model6.fit(X_train, y_train,  
                          epochs = 10, 
                          validation_split = 0.20, 
                          # class_weight = classWeight,
                          batch_size = 64,
                          callbacks = [checkpoint_cb, early_stopping_cb, lr_scheduler])

Epoch 1/10
50/50 [==============================] - 10s 127ms/step - loss: 2.6552 - accuracy: 0.5354 - precision: 0.5784 - recall: 0.5660 - AUC: 0.7512 - val_loss: 1.0427 - val_accuracy: 0.6361 - val_precision: 0.6419 - val_recall: 0.6158 - val_AUC: 0.8194 - lr: 0.0100
Epoch 2/10
50/50 [==============================] - 4s 82ms/step - loss: 0.8501 - accuracy: 0.6809 - precision: 0.6911 - recall: 0.6640 - AUC: 0.8532 - val_loss: 1.1066 - val_accuracy: 0.6412 - val_precision: 0.6551 - val_recall: 0.6260 - val_AUC: 0.8166 - lr: 0.0089
Epoch 3/10
50/50 [==============================] - 5s 93ms/step - loss: 0.8175 - accuracy: 0.6748 - precision: 0.6856 - recall: 0.6624 - AUC: 0.8605 - val_loss: 0.8326 - val_accuracy: 0.7010 - val_precision: 0.7120 - val_recall: 0.6794 - val_AUC: 0.8631 - lr: 0.0079
Epoch 4/10
50/50 [==============================] - 7s 140ms/step - loss: 0.5803 - accuracy: 0.7646 - precision: 0.7810 - recall: 0.7462 - AUC: 0.9130 - val_loss: 0.7398 - val_accuracy: 0.7226 - val_precision: 0.7312 - val_recall: 0.7061 - val_AUC: 0.8748 - lr: 0.0071
Epoch 5/10
50/50 [==============================] - 5s 91ms/step - loss: 0.5684 - accuracy: 0.7586 - precision: 0.7719 - recall: 0.7414 - AUC: 0.9150 - val_loss: 1.0710 - val_accuracy: 0.6158 - val_precision: 0.6203 - val_recall: 0.6132 - val_AUC: 0.8038 - lr: 0.0063
Epoch 6/10
50/50 [==============================] - 4s 79ms/step - loss: 0.4736 - accuracy: 0.8032 - precision: 0.8195 - recall: 0.7854 - AUC: 0.9380 - val_loss: 0.7586 - val_accuracy: 0.7417 - val_precision: 0.7493 - val_recall: 0.7188 - val_AUC: 0.8741 - lr: 0.0056
Epoch 7/10
50/50 [==============================] - 4s 80ms/step - loss: 0.4623 - accuracy: 0.8064 - precision: 0.8191 - recall: 0.7914 - AUC: 0.9409 - val_loss: 0.8553 - val_accuracy: 0.6883 - val_precision: 0.7087 - val_recall: 0.6718 - val_AUC: 0.8650 - lr: 0.0050
Epoch 8/10
50/50 [==============================] - 5s 91ms/step - loss: 0.3989 - accuracy: 0.8303 - precision: 0.8481 - recall: 0.8159 - AUC: 0.9560 - val_loss: 0.9018 - val_accuracy: 0.6832 - val_precision: 0.7020 - val_recall: 0.6654 - val_AUC: 0.8580 - lr: 0.0045
Epoch 9/10
50/50 [==============================] - 5s 92ms/step - loss: 0.3871 - accuracy: 0.8484 - precision: 0.8618 - recall: 0.8322 - AUC: 0.9587 - val_loss: 0.7575 - val_accuracy: 0.7443 - val_precision: 0.7539 - val_recall: 0.7328 - val_AUC: 0.8772 - lr: 0.0040
time: 48 s (started: 2023-03-25 02:10:01 +00:00)


# Show results and print graphs
results = model6.evaluate(X_test, y_test)
results = dict(zip(model6.metrics_names,results))

print(results)
plottingScores(H)

53/53 [==============================] - 3s 43ms/step - loss: 0.6947 - accuracy: 0.6952 - precision: 0.7083 - recall: 0.6768 - AUC: 0.8793
{'loss': 0.6946563720703125, 'accuracy': 0.6951871514320374, 'precision': 0.7083333134651184, 'recall': 0.6767676472663879, 'AUC': 0.8792810440063477}

time: 5.44 s (started: 2023-03-25 02:10:49 +00:00)


y_pred = model6.predict(X_test)
# y_pred = np.array([0 if proba < 0.5 else 1 for proba in y_pred_proba])

53/53 [==============================] - 3s 32ms/step
time: 3.25 s (started: 2023-03-25 02:10:54 +00:00)


# Classification Report
print("Classification Report:\n",classification_report(y_test.argmax(axis=1), y_pred.argmax(axis=1)))

# Confusion Matrix
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test.argmax(axis=1), y_pred.argmax(axis=1))
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1', '2']],  
                         columns = [i for i in ['0', '1', '2']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.76      0.75      0.76       603
           1       0.59      0.64      0.62       600
           2       0.75      0.69      0.72       480

    accuracy                           0.70      1683
   macro avg       0.70      0.69      0.70      1683
weighted avg       0.70      0.70      0.70      1683

Confusion Matrix Chart:

time: 331 ms (started: 2023-03-25 02:10:57 +00:00)


# Model comparison
precision = precision_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
recall = recall_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
f1 = f1_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
AUC = results['AUC']

Train_Accuracy = model6.evaluate(X_train, y_train)
Test_Accuracy = model6.evaluate(X_test, y_test)

# base_1 = []
base_1.append(['Model-6: ResNet50', Train_Accuracy[1], Test_Accuracy[1], precision, recall, f1, AUC])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

123/123 [==============================] - 5s 38ms/step - loss: 0.4981 - accuracy: 0.7990 - precision: 0.8134 - recall: 0.7815 - AUC: 0.9330
53/53 [==============================] - 2s 35ms/step - loss: 0.6947 - accuracy: 0.6952 - precision: 0.7083 - recall: 0.6768 - AUC: 0.8793
time: 7.85 s (started: 2023-03-25 02:10:58 +00:00)


# Function to create NN with transfer learning
def enetv2b3_model():
    model = tf.keras.Sequential([
        tf.keras.applications.EfficientNetV2B3(include_top = False, weights="imagenet", input_shape=(128, 128, 3)),
        tf.keras.layers.GlobalAveragePooling2D(),
        Dense(3, activation = 'softmax')])
    
    model.layers[0].trainable = False
    
    return model

time: 1.1 ms (started: 2023-03-25 02:11:06 +00:00)


# Build and compile the model
model7 = enetv2b3_model()
model7.summary()
model7.compile(optimizer='adam', loss='categorical_crossentropy', metrics=METRICS)

Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/efficientnet_v2/efficientnetv2-b3_notop.h5
52606240/52606240 [==============================] - 3s 0us/step
Model: "sequential_6"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 efficientnetv2-b3 (Function  (None, 4, 4, 1536)       12930622  
 al)                                                             
                                                                 
 global_average_pooling2d_3   (None, 1536)             0         
 (GlobalAveragePooling2D)                                        
                                                                 
 dense_10 (Dense)            (None, 3)                 4611      
                                                                 
=================================================================
Total params: 12,935,233
Trainable params: 4,611
Non-trainable params: 12,930,622
_________________________________________________________________
time: 8.83 s (started: 2023-03-25 02:11:06 +00:00)


H = model7.fit(X_train, y_train,  
                          epochs = 10, 
                          validation_split = 0.20, 
                          # class_weight = classWeight,
                          batch_size = 64,
                          callbacks = [checkpoint_cb, early_stopping_cb, lr_scheduler])

Epoch 1/10
50/50 [==============================] - 21s 175ms/step - loss: 0.9627 - accuracy: 0.5761 - precision: 0.6403 - recall: 0.5847 - AUC: 0.8156 - val_loss: 0.8842 - val_accuracy: 0.6374 - val_precision: 0.6662 - val_recall: 0.5941 - val_AUC: 0.8146 - lr: 0.0100
Epoch 2/10
50/50 [==============================] - 6s 125ms/step - loss: 0.8006 - accuracy: 0.6401 - precision: 0.6625 - recall: 0.6057 - AUC: 0.8293 - val_loss: 0.7372 - val_accuracy: 0.6807 - val_precision: 0.7093 - val_recall: 0.6425 - val_AUC: 0.8570 - lr: 0.0089
Epoch 3/10
50/50 [==============================] - 4s 87ms/step - loss: 0.7405 - accuracy: 0.6736 - precision: 0.6988 - recall: 0.6369 - AUC: 0.8500 - val_loss: 0.8033 - val_accuracy: 0.6807 - val_precision: 0.6990 - val_recall: 0.6501 - val_AUC: 0.8494 - lr: 0.0079
Epoch 4/10
50/50 [==============================] - 4s 77ms/step - loss: 0.7226 - accuracy: 0.6748 - precision: 0.7007 - recall: 0.6398 - AUC: 0.8576 - val_loss: 0.7808 - val_accuracy: 0.6578 - val_precision: 0.6763 - val_recall: 0.6272 - val_AUC: 0.8359 - lr: 0.0071
Epoch 5/10
50/50 [==============================] - 4s 87ms/step - loss: 0.7202 - accuracy: 0.6847 - precision: 0.7037 - recall: 0.6408 - AUC: 0.8569 - val_loss: 0.7512 - val_accuracy: 0.6896 - val_precision: 0.7070 - val_recall: 0.6539 - val_AUC: 0.8576 - lr: 0.0063
Epoch 6/10
50/50 [==============================] - 4s 86ms/step - loss: 0.6824 - accuracy: 0.7045 - precision: 0.7205 - recall: 0.6691 - AUC: 0.8708 - val_loss: 0.7733 - val_accuracy: 0.6616 - val_precision: 0.6702 - val_recall: 0.6412 - val_AUC: 0.8428 - lr: 0.0056
Epoch 7/10
50/50 [==============================] - 6s 126ms/step - loss: 0.6565 - accuracy: 0.7217 - precision: 0.7451 - recall: 0.6850 - AUC: 0.8801 - val_loss: 0.7055 - val_accuracy: 0.6921 - val_precision: 0.7151 - val_recall: 0.6578 - val_AUC: 0.8704 - lr: 0.0050
Epoch 8/10
50/50 [==============================] - 5s 106ms/step - loss: 0.6377 - accuracy: 0.7137 - precision: 0.7388 - recall: 0.6755 - AUC: 0.8861 - val_loss: 0.6972 - val_accuracy: 0.6947 - val_precision: 0.7149 - val_recall: 0.6603 - val_AUC: 0.8686 - lr: 0.0045
Epoch 9/10
50/50 [==============================] - 5s 95ms/step - loss: 0.6267 - accuracy: 0.7194 - precision: 0.7411 - recall: 0.6809 - AUC: 0.8896 - val_loss: 0.6773 - val_accuracy: 0.7188 - val_precision: 0.7418 - val_recall: 0.6870 - val_AUC: 0.8773 - lr: 0.0040
Epoch 10/10
50/50 [==============================] - 4s 90ms/step - loss: 0.6192 - accuracy: 0.7283 - precision: 0.7519 - recall: 0.6949 - AUC: 0.8921 - val_loss: 0.6964 - val_accuracy: 0.7061 - val_precision: 0.7199 - val_recall: 0.6705 - val_AUC: 0.8720 - lr: 0.0035
time: 1min 5s (started: 2023-03-25 02:11:14 +00:00)


# Show results and print graphs
results = model7.evaluate(X_test, y_test)
results = dict(zip(model7.metrics_names,results))

print(results)
plottingScores(H)

53/53 [==============================] - 3s 43ms/step - loss: 0.6793 - accuracy: 0.6934 - precision: 0.7138 - recall: 0.6684 - AUC: 0.8696
{'loss': 0.6793166995048523, 'accuracy': 0.6934046149253845, 'precision': 0.7138324975967407, 'recall': 0.6684492230415344, 'AUC': 0.8695988655090332}

time: 6.88 s (started: 2023-03-25 02:12:20 +00:00)


y_pred = model7.predict(X_test)
# y_pred = np.array([0 if proba < 0.5 else 1 for proba in y_pred_proba])

53/53 [==============================] - 4s 35ms/step
time: 5.07 s (started: 2023-03-25 02:12:27 +00:00)


# Classification Report
print("Classification Report:\n",classification_report(y_test.argmax(axis=1), y_pred.argmax(axis=1)))

# Confusion Matrix
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test.argmax(axis=1), y_pred.argmax(axis=1))
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1', '2']],  
                         columns = [i for i in ['0', '1', '2']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.77      0.70      0.73       603
           1       0.61      0.58      0.60       600
           2       0.71      0.82      0.76       480

    accuracy                           0.69      1683
   macro avg       0.69      0.70      0.70      1683
weighted avg       0.69      0.69      0.69      1683

Confusion Matrix Chart:

time: 500 ms (started: 2023-03-25 02:12:32 +00:00)


# Model comparison
precision = precision_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
recall = recall_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
f1 = f1_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
AUC = results['AUC']

Train_Accuracy = model7.evaluate(X_train, y_train)
Test_Accuracy = model7.evaluate(X_test, y_test)

# base_1 = []
base_1.append(['Model-7: EfficientNetV2B3', Train_Accuracy[1], Test_Accuracy[1], precision, recall, f1, AUC])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

123/123 [==============================] - 5s 38ms/step - loss: 0.5394 - accuracy: 0.7725 - precision: 0.7990 - recall: 0.7481 - AUC: 0.9221
53/53 [==============================] - 2s 35ms/step - loss: 0.6793 - accuracy: 0.6934 - precision: 0.7138 - recall: 0.6684 - AUC: 0.8696
time: 8.29 s (started: 2023-03-25 02:12:32 +00:00)


# # Function to create and compile model, required for KerasClassifier
def cnn_model(lr, batch_size):
    model = tf.keras.Sequential([
    tf.keras.layers.Conv2D(32, kernel_size=(3,3), strides=(1,1), padding = 'valid', activation = 'relu', input_shape=(128, 128, 3)), # convolutional layer
    tf.keras.layers.MaxPool2D(pool_size=(2,2)), # flatten output of conv
        
    tf.keras.layers.Conv2D(32, kernel_size=(3,3), strides=(1,1), padding = 'valid', activation = 'relu'), # convolutional layer
    tf.keras.layers.MaxPool2D(pool_size=(2,2)), # flatten output of conv
    tf.keras.layers.Dropout(0.3),
        
    tf.keras.layers.Conv2D(64, 3, activation = 'relu', padding = 'valid'),
    tf.keras.layers.Conv2D(128, 3, activation = 'relu', padding = 'valid'),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.MaxPool2D(),
    tf.keras.layers.Dropout(0.4),

    tf.keras.layers.Flatten(), # flatten output of conv
    tf.keras.layers.Dense(512, activation = "relu"), # hidden layer
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.5),
    tf.keras.layers.Dense(128, activation = "relu"), #  output layer
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.5),
    tf.keras.layers.Dense(3, activation = "softmax")])
    # compile the model
    opt = tf.keras.optimizers.Adam(learning_rate=lr)
    model.compile(optimizer=opt, loss="categorical_crossentropy", metrics=METRICS)  
    return model

time: 2.97 ms (started: 2023-03-25 02:12:41 +00:00)


# Create the model
model = KerasClassifier(build_fn = cnn_model, verbose=1)

# Grid search for batch size and learning rate
params = {'batch_size':[32, 64, 128],
          'lr':[0.01,0.1,0.001]}
          
# Grid search for batch size, learning rate, optimizers, activation, epochs, layers, neurons, etc.
# params = {'batch_size':[32, 64, 128],
#           'optimizers':['rmsprop', 'adam', 'SGD'],
#           'activation':['relu', 'sigmoid', 'tanh'],
#           'epochs':[50, 100, 150],
#           'lr':[0.01, 0.1, 0.001]}

gs = RandomizedSearchCV(estimator=model, param_distributions=params, cv=5, verbose=10, random_state=0) # For local machine n_jobs=2
gs.fit(X_train, y_train)

Fitting 5 folds for each of 9 candidates, totalling 45 fits
[CV 1/5; 1/9] START batch_size=32, lr=0.01......................................
99/99 [==============================] - 8s 25ms/step - loss: 1.1653 - accuracy: 0.4873 - precision: 0.5861 - recall: 0.4935 - AUC: 0.7523
25/25 [==============================] - 1s 7ms/step - loss: 5.3441 - accuracy: 0.3295 - precision: 0.3277 - recall: 0.3193 - AUC: 0.5259
[CV 1/5; 1/9] END .......batch_size=32, lr=0.01;, score=0.330 total time=   9.0s
[CV 2/5; 1/9] START batch_size=32, lr=0.01......................................
99/99 [==============================] - 6s 25ms/step - loss: 1.1573 - accuracy: 0.4976 - precision: 0.4699 - recall: 0.3873 - AUC: 0.6262
25/25 [==============================] - 1s 13ms/step - loss: 4.0210 - accuracy: 0.2981 - precision: 0.2978 - recall: 0.2955 - AUC: 0.5210
[CV 2/5; 1/9] END .......batch_size=32, lr=0.01;, score=0.298 total time=   8.0s
[CV 3/5; 1/9] START batch_size=32, lr=0.01......................................
99/99 [==============================] - 6s 24ms/step - loss: 1.2062 - accuracy: 0.4922 - precision: 0.4558 - recall: 0.3734 - AUC: 0.6130
25/25 [==============================] - 1s 8ms/step - loss: 2.5138 - accuracy: 0.4166 - precision: 0.4223 - recall: 0.3809 - AUC: 0.6022
[CV 3/5; 1/9] END .......batch_size=32, lr=0.01;, score=0.417 total time=   7.9s
[CV 4/5; 1/9] START batch_size=32, lr=0.01......................................
99/99 [==============================] - 6s 25ms/step - loss: 1.1749 - accuracy: 0.4858 - precision: 0.4895 - recall: 0.3961 - AUC: 0.6515
25/25 [==============================] - 1s 11ms/step - loss: 4.3997 - accuracy: 0.3758 - precision: 0.3758 - recall: 0.3758 - AUC: 0.5727
[CV 4/5; 1/9] END .......batch_size=32, lr=0.01;, score=0.376 total time=   7.5s
[CV 5/5; 1/9] START batch_size=32, lr=0.01......................................
99/99 [==============================] - 6s 24ms/step - loss: 1.1693 - accuracy: 0.5059 - precision: 0.4854 - recall: 0.4121 - AUC: 0.6427
25/25 [==============================] - 1s 7ms/step - loss: 8.2771 - accuracy: 0.3083 - precision: 0.3065 - recall: 0.3057 - AUC: 0.4794
[CV 5/5; 1/9] END .......batch_size=32, lr=0.01;, score=0.308 total time=   8.6s
[CV 1/5; 2/9] START batch_size=32, lr=0.1.......................................
99/99 [==============================] - 7s 26ms/step - loss: 1.5493 - accuracy: 0.4513 - precision: 0.4250 - recall: 0.3702 - AUC: 0.5902
25/25 [==============================] - 1s 8ms/step - loss: 5.8272 - accuracy: 0.3473 - precision: 0.3484 - recall: 0.3435 - AUC: 0.5223
[CV 1/5; 2/9] END ........batch_size=32, lr=0.1;, score=0.347 total time=  13.3s
[CV 2/5; 2/9] START batch_size=32, lr=0.1.......................................
99/99 [==============================] - 7s 27ms/step - loss: 1.5231 - accuracy: 0.4667 - precision: 0.4489 - recall: 0.3911 - AUC: 0.6130
25/25 [==============================] - 1s 7ms/step - loss: 3.4624 - accuracy: 0.4866 - precision: 0.4908 - recall: 0.4739 - AUC: 0.6595
[CV 2/5; 2/9] END ........batch_size=32, lr=0.1;, score=0.487 total time=   8.1s
[CV 3/5; 2/9] START batch_size=32, lr=0.1.......................................
99/99 [==============================] - 6s 23ms/step - loss: 1.5077 - accuracy: 0.4562 - precision: 0.4764 - recall: 0.4017 - AUC: 0.6358
25/25 [==============================] - 1s 7ms/step - loss: 8.3106 - accuracy: 0.2586 - precision: 0.2586 - recall: 0.2586 - AUC: 0.4478
[CV 3/5; 2/9] END ........batch_size=32, lr=0.1;, score=0.259 total time=   7.1s
[CV 4/5; 2/9] START batch_size=32, lr=0.1.......................................
99/99 [==============================] - 8s 28ms/step - loss: 1.6341 - accuracy: 0.4530 - precision: 0.4188 - recall: 0.3556 - AUC: 0.5744
25/25 [==============================] - 1s 8ms/step - loss: 2.3829 - accuracy: 0.4306 - precision: 0.4360 - recall: 0.3860 - AUC: 0.6142
[CV 4/5; 2/9] END ........batch_size=32, lr=0.1;, score=0.431 total time=  14.7s
[CV 5/5; 2/9] START batch_size=32, lr=0.1.......................................
99/99 [==============================] - 7s 25ms/step - loss: 1.6149 - accuracy: 0.4448 - precision: 0.4491 - recall: 0.3821 - AUC: 0.6246
25/25 [==============================] - 1s 8ms/step - loss: 9.4319 - accuracy: 0.4408 - precision: 0.4291 - recall: 0.4089 - AUC: 0.6455
[CV 5/5; 2/9] END ........batch_size=32, lr=0.1;, score=0.441 total time=   8.4s
[CV 1/5; 3/9] START batch_size=32, lr=0.001.....................................
99/99 [==============================] - 6s 24ms/step - loss: 1.5195 - accuracy: 0.4627 - precision: 0.4618 - recall: 0.4161 - AUC: 0.6424
25/25 [==============================] - 1s 8ms/step - loss: 2.4080 - accuracy: 0.4275 - precision: 0.4273 - recall: 0.4262 - AUC: 0.6297
[CV 1/5; 3/9] END ......batch_size=32, lr=0.001;, score=0.427 total time=   7.3s
[CV 2/5; 3/9] START batch_size=32, lr=0.001.....................................
99/99 [==============================] - 7s 25ms/step - loss: 1.4466 - accuracy: 0.4865 - precision: 0.4833 - recall: 0.4431 - AUC: 0.6474
25/25 [==============================] - 1s 8ms/step - loss: 3.8638 - accuracy: 0.3898 - precision: 0.3903 - recall: 0.3898 - AUC: 0.5660
[CV 2/5; 3/9] END ......batch_size=32, lr=0.001;, score=0.390 total time=   9.0s
[CV 3/5; 3/9] START batch_size=32, lr=0.001.....................................
99/99 [==============================] - 6s 26ms/step - loss: 1.4415 - accuracy: 0.4572 - precision: 0.4564 - recall: 0.4162 - AUC: 0.6276
25/25 [==============================] - 1s 9ms/step - loss: 1.4605 - accuracy: 0.5108 - precision: 0.5201 - recall: 0.4955 - AUC: 0.7119
[CV 3/5; 3/9] END ......batch_size=32, lr=0.001;, score=0.511 total time=   8.4s
[CV 4/5; 3/9] START batch_size=32, lr=0.001.....................................
99/99 [==============================] - 6s 25ms/step - loss: 1.4697 - accuracy: 0.4432 - precision: 0.4642 - recall: 0.4142 - AUC: 0.6507
25/25 [==============================] - 1s 8ms/step - loss: 3.6239 - accuracy: 0.3707 - precision: 0.3686 - recall: 0.3643 - AUC: 0.5751
[CV 4/5; 3/9] END ......batch_size=32, lr=0.001;, score=0.371 total time=   7.2s
[CV 5/5; 3/9] START batch_size=32, lr=0.001.....................................
99/99 [==============================] - 7s 29ms/step - loss: 1.4369 - accuracy: 0.4632 - precision: 0.4507 - recall: 0.4065 - AUC: 0.6170
25/25 [==============================] - 1s 8ms/step - loss: 3.0144 - accuracy: 0.3669 - precision: 0.3713 - recall: 0.3656 - AUC: 0.5574
[CV 5/5; 3/9] END ......batch_size=32, lr=0.001;, score=0.367 total time=   8.2s
[CV 1/5; 4/9] START batch_size=64, lr=0.01......................................
50/50 [==============================] - 6s 40ms/step - loss: 1.2157 - accuracy: 0.4828 - precision: 0.4705 - recall: 0.3860 - AUC: 0.6333
13/13 [==============================] - 1s 11ms/step - loss: 12.4178 - accuracy: 0.3435 - precision: 0.3435 - recall: 0.3435 - AUC: 0.5123
[CV 1/5; 4/9] END .......batch_size=64, lr=0.01;, score=0.344 total time=   7.0s
[CV 2/5; 4/9] START batch_size=64, lr=0.01......................................
50/50 [==============================] - 7s 42ms/step - loss: 1.2077 - accuracy: 0.4954 - precision: 0.4785 - recall: 0.4057 - AUC: 0.6216
13/13 [==============================] - 1s 12ms/step - loss: 26.1691 - accuracy: 0.3490 - precision: 0.3490 - recall: 0.3490 - AUC: 0.5118
[CV 2/5; 4/9] END .......batch_size=64, lr=0.01;, score=0.349 total time=   8.6s
[CV 3/5; 4/9] START batch_size=64, lr=0.01......................................
50/50 [==============================] - 7s 41ms/step - loss: 1.2212 - accuracy: 0.4833 - precision: 0.4654 - recall: 0.3884 - AUC: 0.6116
13/13 [==============================] - 1s 12ms/step - loss: 10.9771 - accuracy: 0.2713 - precision: 0.2713 - recall: 0.2713 - AUC: 0.4681
[CV 3/5; 4/9] END .......batch_size=64, lr=0.01;, score=0.271 total time=   8.3s
[CV 4/5; 4/9] START batch_size=64, lr=0.01......................................
50/50 [==============================] - 6s 41ms/step - loss: 1.2156 - accuracy: 0.4760 - precision: 0.4463 - recall: 0.3800 - AUC: 0.5961
13/13 [==============================] - 0s 10ms/step - loss: 15.9224 - accuracy: 0.3643 - precision: 0.3643 - recall: 0.3643 - AUC: 0.5228
[CV 4/5; 4/9] END .......batch_size=64, lr=0.01;, score=0.364 total time=   7.4s
[CV 5/5; 4/9] START batch_size=64, lr=0.01......................................
50/50 [==============================] - 6s 42ms/step - loss: 1.1881 - accuracy: 0.4804 - precision: 0.4658 - recall: 0.3816 - AUC: 0.6180
13/13 [==============================] - 1s 13ms/step - loss: 8.1534 - accuracy: 0.4127 - precision: 0.4127 - recall: 0.4127 - AUC: 0.5656
[CV 5/5; 4/9] END .......batch_size=64, lr=0.01;, score=0.413 total time=   8.3s
[CV 1/5; 5/9] START batch_size=64, lr=0.1.......................................
50/50 [==============================] - 5s 41ms/step - loss: 1.6474 - accuracy: 0.4449 - precision: 0.4432 - recall: 0.3949 - AUC: 0.6093
13/13 [==============================] - 1s 12ms/step - loss: 64.3442 - accuracy: 0.3664 - precision: 0.3664 - recall: 0.3664 - AUC: 0.5248
[CV 1/5; 5/9] END ........batch_size=64, lr=0.1;, score=0.366 total time=   8.2s
[CV 2/5; 5/9] START batch_size=64, lr=0.1.......................................
50/50 [==============================] - 7s 41ms/step - loss: 1.5177 - accuracy: 0.4473 - precision: 0.4357 - recall: 0.3838 - AUC: 0.5976
13/13 [==============================] - 0s 11ms/step - loss: 26.5604 - accuracy: 0.2777 - precision: 0.2777 - recall: 0.2777 - AUC: 0.4688
[CV 2/5; 5/9] END ........batch_size=64, lr=0.1;, score=0.278 total time=   8.3s
[CV 3/5; 5/9] START batch_size=64, lr=0.1.......................................
50/50 [==============================] - 6s 42ms/step - loss: 1.6750 - accuracy: 0.4543 - precision: 0.4222 - recall: 0.3739 - AUC: 0.5742
13/13 [==============================] - 1s 15ms/step - loss: 21.1854 - accuracy: 0.3771 - precision: 0.3771 - recall: 0.3771 - AUC: 0.5328
[CV 3/5; 5/9] END ........batch_size=64, lr=0.1;, score=0.377 total time=   8.5s
[CV 4/5; 5/9] START batch_size=64, lr=0.1.......................................
50/50 [==============================] - 6s 42ms/step - loss: 1.5214 - accuracy: 0.4387 - precision: 0.4324 - recall: 0.3683 - AUC: 0.5981
13/13 [==============================] - 1s 13ms/step - loss: 51.2503 - accuracy: 0.3172 - precision: 0.3172 - recall: 0.3172 - AUC: 0.4862
[CV 4/5; 5/9] END ........batch_size=64, lr=0.1;, score=0.317 total time=   8.8s
[CV 5/5; 5/9] START batch_size=64, lr=0.1.......................................
50/50 [==============================] - 6s 45ms/step - loss: 1.6315 - accuracy: 0.4438 - precision: 0.4169 - recall: 0.3704 - AUC: 0.5783
13/13 [==============================] - 1s 11ms/step - loss: 19.9058 - accuracy: 0.4051 - precision: 0.4051 - recall: 0.4051 - AUC: 0.5630
[CV 5/5; 5/9] END ........batch_size=64, lr=0.1;, score=0.405 total time=   8.3s
[CV 1/5; 6/9] START batch_size=64, lr=0.001.....................................
50/50 [==============================] - 6s 42ms/step - loss: 1.3862 - accuracy: 0.4682 - precision: 0.4696 - recall: 0.4285 - AUC: 0.6291
13/13 [==============================] - 1s 11ms/step - loss: 31.4228 - accuracy: 0.2850 - precision: 0.2850 - recall: 0.2850 - AUC: 0.4636
[CV 1/5; 6/9] END ......batch_size=64, lr=0.001;, score=0.285 total time=   7.0s
[CV 2/5; 6/9] START batch_size=64, lr=0.001.....................................
50/50 [==============================] - 8s 44ms/step - loss: 1.4638 - accuracy: 0.4658 - precision: 0.4330 - recall: 0.3972 - AUC: 0.5906
13/13 [==============================] - 1s 14ms/step - loss: 1.6801 - accuracy: 0.4637 - precision: 0.4667 - recall: 0.4548 - AUC: 0.6776
[CV 2/5; 6/9] END ......batch_size=64, lr=0.001;, score=0.464 total time=   9.9s
[CV 3/5; 6/9] START batch_size=64, lr=0.001.....................................
50/50 [==============================] - 6s 42ms/step - loss: 1.4399 - accuracy: 0.4807 - precision: 0.4856 - recall: 0.4376 - AUC: 0.6602
13/13 [==============================] - 1s 14ms/step - loss: 3.4242 - accuracy: 0.3745 - precision: 0.3742 - recall: 0.3694 - AUC: 0.5740
[CV 3/5; 6/9] END ......batch_size=64, lr=0.001;, score=0.375 total time=   8.8s
[CV 4/5; 6/9] START batch_size=64, lr=0.001.....................................
50/50 [==============================] - 6s 43ms/step - loss: 1.4769 - accuracy: 0.4559 - precision: 0.4462 - recall: 0.4096 - AUC: 0.6194
13/13 [==============================] - 1s 12ms/step - loss: 4.3765 - accuracy: 0.3669 - precision: 0.3627 - recall: 0.3567 - AUC: 0.5642
[CV 4/5; 6/9] END ......batch_size=64, lr=0.001;, score=0.367 total time=   8.3s
[CV 5/5; 6/9] START batch_size=64, lr=0.001.....................................
50/50 [==============================] - 7s 43ms/step - loss: 1.5237 - accuracy: 0.4623 - precision: 0.4480 - recall: 0.4134 - AUC: 0.6143
13/13 [==============================] - 1s 11ms/step - loss: 11.6339 - accuracy: 0.4344 - precision: 0.4344 - recall: 0.4344 - AUC: 0.5687
[CV 5/5; 6/9] END ......batch_size=64, lr=0.001;, score=0.434 total time=   8.6s
[CV 1/5; 7/9] START batch_size=128, lr=0.01.....................................
25/25 [==============================] - 5s 74ms/step - loss: 1.3844 - accuracy: 0.4576 - precision: 0.4668 - recall: 0.3997 - AUC: 0.6225
7/7 [==============================] - 1s 23ms/step - loss: 83.3901 - accuracy: 0.3715 - precision: 0.3715 - recall: 0.3715 - AUC: 0.5286
[CV 1/5; 7/9] END ......batch_size=128, lr=0.01;, score=0.372 total time=   8.5s
[CV 2/5; 7/9] START batch_size=128, lr=0.01.....................................
25/25 [==============================] - 6s 94ms/step - loss: 1.2110 - accuracy: 0.5018 - precision: 0.4836 - recall: 0.4156 - AUC: 0.6276
7/7 [==============================] - 0s 15ms/step - loss: 123.8430 - accuracy: 0.2713 - precision: 0.2713 - recall: 0.2713 - AUC: 0.4535
[CV 2/5; 7/9] END ......batch_size=128, lr=0.01;, score=0.271 total time=   7.4s
[CV 3/5; 7/9] START batch_size=128, lr=0.01.....................................
25/25 [==============================] - 5s 79ms/step - loss: 1.2283 - accuracy: 0.4970 - precision: 0.4621 - recall: 0.3895 - AUC: 0.6025
7/7 [==============================] - 1s 20ms/step - loss: 88.0434 - accuracy: 0.2611 - precision: 0.2611 - recall: 0.2611 - AUC: 0.4452
[CV 3/5; 7/9] END ......batch_size=128, lr=0.01;, score=0.261 total time=   8.1s
[CV 4/5; 7/9] START batch_size=128, lr=0.01.....................................
25/25 [==============================] - 5s 77ms/step - loss: 1.2939 - accuracy: 0.4807 - precision: 0.4464 - recall: 0.3839 - AUC: 0.5873
7/7 [==============================] - 0s 16ms/step - loss: 105.9672 - accuracy: 0.3261 - precision: 0.3261 - recall: 0.3261 - AUC: 0.4959
[CV 4/5; 7/9] END ......batch_size=128, lr=0.01;, score=0.326 total time=   6.5s
[CV 5/5; 7/9] START batch_size=128, lr=0.01.....................................
25/25 [==============================] - 6s 81ms/step - loss: 1.2657 - accuracy: 0.4702 - precision: 0.4521 - recall: 0.3874 - AUC: 0.6091
7/7 [==============================] - 1s 17ms/step - loss: 102.9465 - accuracy: 0.3019 - precision: 0.3019 - recall: 0.3019 - AUC: 0.4764
[CV 5/5; 7/9] END ......batch_size=128, lr=0.01;, score=0.302 total time=   7.4s
[CV 1/5; 8/9] START batch_size=128, lr=0.1......................................
25/25 [==============================] - 6s 75ms/step - loss: 1.6077 - accuracy: 0.4475 - precision: 0.4262 - recall: 0.3761 - AUC: 0.5761
7/7 [==============================] - 1s 25ms/step - loss: 80.3909 - accuracy: 0.2837 - precision: 0.2837 - recall: 0.2837 - AUC: 0.4628
[CV 1/5; 8/9] END .......batch_size=128, lr=0.1;, score=0.284 total time=   9.7s
[CV 2/5; 8/9] START batch_size=128, lr=0.1......................................
25/25 [==============================] - 6s 75ms/step - loss: 1.6053 - accuracy: 0.4502 - precision: 0.4193 - recall: 0.3769 - AUC: 0.5799
7/7 [==============================] - 1s 20ms/step - loss: 40.2148 - accuracy: 0.3452 - precision: 0.3452 - recall: 0.3452 - AUC: 0.5061
[CV 2/5; 8/9] END .......batch_size=128, lr=0.1;, score=0.345 total time=   9.3s
[CV 3/5; 8/9] START batch_size=128, lr=0.1......................................
25/25 [==============================] - 6s 80ms/step - loss: 1.6123 - accuracy: 0.4648 - precision: 0.4430 - recall: 0.3951 - AUC: 0.5934
7/7 [==============================] - 1s 20ms/step - loss: 78.8507 - accuracy: 0.2650 - precision: 0.2650 - recall: 0.2650 - AUC: 0.4486
[CV 3/5; 8/9] END .......batch_size=128, lr=0.1;, score=0.265 total time=   7.0s
[CV 4/5; 8/9] START batch_size=128, lr=0.1......................................
25/25 [==============================] - 6s 76ms/step - loss: 1.7243 - accuracy: 0.4444 - precision: 0.4151 - recall: 0.3749 - AUC: 0.5640
7/7 [==============================] - 0s 16ms/step - loss: 229.5545 - accuracy: 0.3631 - precision: 0.3631 - recall: 0.3631 - AUC: 0.5223
[CV 4/5; 8/9] END .......batch_size=128, lr=0.1;, score=0.363 total time=   6.8s
[CV 5/5; 8/9] START batch_size=128, lr=0.1......................................
25/25 [==============================] - 5s 79ms/step - loss: 1.6140 - accuracy: 0.4298 - precision: 0.4243 - recall: 0.3688 - AUC: 0.5834
7/7 [==============================] - 1s 20ms/step - loss: 434.6324 - accuracy: 0.3210 - precision: 0.3210 - recall: 0.3210 - AUC: 0.4908
[CV 5/5; 8/9] END .......batch_size=128, lr=0.1;, score=0.321 total time=   8.3s
[CV 1/5; 9/9] START batch_size=128, lr=0.001....................................
25/25 [==============================] - 5s 78ms/step - loss: 1.5310 - accuracy: 0.4717 - precision: 0.4459 - recall: 0.4102 - AUC: 0.5889
7/7 [==============================] - 1s 22ms/step - loss: 9.8206 - accuracy: 0.4326 - precision: 0.4326 - recall: 0.4326 - AUC: 0.5865
[CV 1/5; 9/9] END .....batch_size=128, lr=0.001;, score=0.433 total time=   8.1s
[CV 2/5; 9/9] START batch_size=128, lr=0.001....................................
25/25 [==============================] - 7s 79ms/step - loss: 1.4713 - accuracy: 0.4788 - precision: 0.4770 - recall: 0.4433 - AUC: 0.6376
7/7 [==============================] - 0s 17ms/step - loss: 15.4613 - accuracy: 0.2790 - precision: 0.2790 - recall: 0.2790 - AUC: 0.4585
[CV 2/5; 9/9] END .....batch_size=128, lr=0.001;, score=0.279 total time=   7.8s
[CV 3/5; 9/9] START batch_size=128, lr=0.001....................................
25/25 [==============================] - 5s 80ms/step - loss: 1.5606 - accuracy: 0.4597 - precision: 0.4218 - recall: 0.3900 - AUC: 0.5750
7/7 [==============================] - 1s 23ms/step - loss: 21.3588 - accuracy: 0.2662 - precision: 0.2662 - recall: 0.2662 - AUC: 0.4473
[CV 3/5; 9/9] END .....batch_size=128, lr=0.001;, score=0.266 total time=   8.3s
[CV 4/5; 9/9] START batch_size=128, lr=0.001....................................
25/25 [==============================] - 6s 80ms/step - loss: 1.5126 - accuracy: 0.4760 - precision: 0.4394 - recall: 0.4052 - AUC: 0.5817
7/7 [==============================] - 2s 20ms/step - loss: 26.4982 - accuracy: 0.4318 - precision: 0.4318 - recall: 0.4318 - AUC: 0.5922
[CV 4/5; 9/9] END .....batch_size=128, lr=0.001;, score=0.432 total time=   8.6s
[CV 5/5; 9/9] START batch_size=128, lr=0.001....................................
25/25 [==============================] - 6s 82ms/step - loss: 1.4947 - accuracy: 0.4693 - precision: 0.4709 - recall: 0.4343 - AUC: 0.6311
7/7 [==============================] - 1s 17ms/step - loss: 12.1501 - accuracy: 0.4115 - precision: 0.4115 - recall: 0.4115 - AUC: 0.5759
[CV 5/5; 9/9] END .....batch_size=128, lr=0.001;, score=0.411 total time=   7.7s
123/123 [==============================] - 7s 27ms/step - loss: 1.3821 - accuracy: 0.4783 - precision: 0.4774 - recall: 0.4330 - AUC: 0.6376

RandomizedSearchCV(cv=5,
                   estimator=<keras.wrappers.scikit_learn.KerasClassifier object at 0x7f93ac686790>,
                   param_distributions={'batch_size': [32, 64, 128],
                                        'lr': [0.01, 0.1, 0.001]},
                   random_state=0, verbose=10)

RandomizedSearchCV(cv=5,
                   estimator=<keras.wrappers.scikit_learn.KerasClassifier object at 0x7f93ac686790>,
                   param_distributions={'batch_size': [32, 64, 128],
                                        'lr': [0.01, 0.1, 0.001]},
                   random_state=0, verbose=10)

<keras.wrappers.scikit_learn.KerasClassifier object at 0x7f93ac686790>

<keras.wrappers.scikit_learn.KerasClassifier object at 0x7f93ac686790>

time: 6min 24s (started: 2023-03-25 02:12:41 +00:00)


# Summarize the results in a dataframe
pd.DataFrame({'param': gs.cv_results_["params"], 
              'score mean': gs.cv_results_["mean_test_score"], 
              'score s.d.': gs.cv_results_["std_test_score"],
              'rank': gs.cv_results_["rank_test_score"]})

time: 20.9 ms (started: 2023-03-25 02:19:05 +00:00)


# Select the best parameters
print('Best Parameters:', gs.best_params_, 'mean score: ', gs.best_score_, sep='\n')

Best Parameters:
{'lr': 0.001, 'batch_size': 32}
mean score: 
0.4131394922733307
time: 4.24 ms (started: 2023-03-25 02:19:05 +00:00)


# Build and compile the model with best params
model8 = cnn_model(lr = gs.best_params_['lr'], batch_size = gs.best_params_['batch_size']) #Change this with best params
model8.summary()

opt = tf.keras.optimizers.Adam(gs.best_params_['lr'])
model8.compile(optimizer=opt, loss="categorical_crossentropy", metrics=METRICS)

Model: "sequential_53"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 conv2d_192 (Conv2D)         (None, 126, 126, 32)      896       
                                                                 
 max_pooling2d_144 (MaxPooli  (None, 63, 63, 32)       0         
 ng2D)                                                           
                                                                 
 conv2d_193 (Conv2D)         (None, 61, 61, 32)        9248      
                                                                 
 max_pooling2d_145 (MaxPooli  (None, 30, 30, 32)       0         
 ng2D)                                                           
                                                                 
 dropout_191 (Dropout)       (None, 30, 30, 32)        0         
                                                                 
 conv2d_194 (Conv2D)         (None, 28, 28, 64)        18496     
                                                                 
 conv2d_195 (Conv2D)         (None, 26, 26, 128)       73856     
                                                                 
 batch_normalization_143 (Ba  (None, 26, 26, 128)      512       
 tchNormalization)                                               
                                                                 
 max_pooling2d_146 (MaxPooli  (None, 13, 13, 128)      0         
 ng2D)                                                           
                                                                 
 dropout_192 (Dropout)       (None, 13, 13, 128)       0         
                                                                 
 flatten_49 (Flatten)        (None, 21632)             0         
                                                                 
 dense_149 (Dense)           (None, 512)               11076096  
                                                                 
 batch_normalization_144 (Ba  (None, 512)              2048      
 tchNormalization)                                               
                                                                 
 dropout_193 (Dropout)       (None, 512)               0         
                                                                 
 dense_150 (Dense)           (None, 128)               65664     
                                                                 
 batch_normalization_145 (Ba  (None, 128)              512       
 tchNormalization)                                               
                                                                 
 dropout_194 (Dropout)       (None, 128)               0         
                                                                 
 dense_151 (Dense)           (None, 3)                 387       
                                                                 
=================================================================
Total params: 11,247,715
Trainable params: 11,246,179
Non-trainable params: 1,536
_________________________________________________________________
time: 291 ms (started: 2023-03-25 02:19:05 +00:00)


# Fit the model
H = model8.fit(X_train, y_train,  
                      epochs=30, 
                      validation_split = 0.20, 
                      batch_size=gs.best_params_['batch_size'], # Change this with best params
                      # class_weight=classWeight,
                      callbacks=[checkpoint_cb, early_stopping_cb, lr_scheduler],
                      verbose=1)

Epoch 1/30
99/99 [==============================] - 9s 33ms/step - loss: 1.1766 - accuracy: 0.4854 - precision: 0.4868 - recall: 0.4173 - AUC: 0.6489 - val_loss: 7.7043 - val_accuracy: 0.3422 - val_precision: 0.3422 - val_recall: 0.3422 - val_AUC: 0.5109 - lr: 0.0100
Epoch 2/30
99/99 [==============================] - 3s 26ms/step - loss: 0.9738 - accuracy: 0.5366 - precision: 0.5631 - recall: 0.4166 - AUC: 0.7270 - val_loss: 1.3918 - val_accuracy: 0.4275 - val_precision: 0.4489 - val_recall: 0.3524 - val_AUC: 0.6388 - lr: 0.0089
Epoch 3/30
99/99 [==============================] - 3s 26ms/step - loss: 0.9381 - accuracy: 0.5398 - precision: 0.5725 - recall: 0.4197 - AUC: 0.7418 - val_loss: 0.9466 - val_accuracy: 0.5445 - val_precision: 0.5963 - val_recall: 0.4453 - val_AUC: 0.7465 - lr: 0.0079
Epoch 4/30
99/99 [==============================] - 3s 27ms/step - loss: 0.9113 - accuracy: 0.5519 - precision: 0.5899 - recall: 0.4261 - AUC: 0.7530 - val_loss: 0.9764 - val_accuracy: 0.5407 - val_precision: 0.5493 - val_recall: 0.4746 - val_AUC: 0.7396 - lr: 0.0071
Epoch 5/30
99/99 [==============================] - 3s 30ms/step - loss: 0.8998 - accuracy: 0.5611 - precision: 0.6048 - recall: 0.4455 - AUC: 0.7607 - val_loss: 0.8685 - val_accuracy: 0.5751 - val_precision: 0.6552 - val_recall: 0.4377 - val_AUC: 0.7766 - lr: 0.0063
Epoch 6/30
99/99 [==============================] - 3s 27ms/step - loss: 0.8843 - accuracy: 0.5541 - precision: 0.6051 - recall: 0.4309 - AUC: 0.7653 - val_loss: 0.8780 - val_accuracy: 0.5903 - val_precision: 0.6615 - val_recall: 0.4924 - val_AUC: 0.7869 - lr: 0.0056
Epoch 7/30
99/99 [==============================] - 3s 26ms/step - loss: 0.8685 - accuracy: 0.5838 - precision: 0.6321 - recall: 0.4624 - AUC: 0.7781 - val_loss: 0.8598 - val_accuracy: 0.5941 - val_precision: 0.6611 - val_recall: 0.4567 - val_AUC: 0.7867 - lr: 0.0050
Epoch 8/30
99/99 [==============================] - 3s 26ms/step - loss: 0.8477 - accuracy: 0.5914 - precision: 0.6457 - recall: 0.4672 - AUC: 0.7882 - val_loss: 0.8818 - val_accuracy: 0.5585 - val_precision: 0.6413 - val_recall: 0.4186 - val_AUC: 0.7681 - lr: 0.0045
Epoch 9/30
99/99 [==============================] - 3s 27ms/step - loss: 0.8507 - accuracy: 0.5901 - precision: 0.6437 - recall: 0.4771 - AUC: 0.7862 - val_loss: 0.8680 - val_accuracy: 0.5814 - val_precision: 0.6485 - val_recall: 0.4835 - val_AUC: 0.7810 - lr: 0.0040
Epoch 10/30
99/99 [==============================] - 3s 31ms/step - loss: 0.8442 - accuracy: 0.5863 - precision: 0.6399 - recall: 0.4764 - AUC: 0.7885 - val_loss: 0.8816 - val_accuracy: 0.6056 - val_precision: 0.6616 - val_recall: 0.5025 - val_AUC: 0.7922 - lr: 0.0035
Epoch 11/30
99/99 [==============================] - 3s 27ms/step - loss: 0.8194 - accuracy: 0.6185 - precision: 0.6623 - recall: 0.5022 - AUC: 0.8033 - val_loss: 0.8475 - val_accuracy: 0.6031 - val_precision: 0.6540 - val_recall: 0.5025 - val_AUC: 0.7969 - lr: 0.0032
Epoch 12/30
99/99 [==============================] - 3s 34ms/step - loss: 0.8192 - accuracy: 0.6131 - precision: 0.6639 - recall: 0.5166 - AUC: 0.8052 - val_loss: 0.8388 - val_accuracy: 0.6158 - val_precision: 0.6751 - val_recall: 0.5102 - val_AUC: 0.8054 - lr: 0.0028
Epoch 13/30
99/99 [==============================] - 3s 34ms/step - loss: 0.8120 - accuracy: 0.6178 - precision: 0.6595 - recall: 0.5108 - AUC: 0.8072 - val_loss: 0.8873 - val_accuracy: 0.6209 - val_precision: 0.6700 - val_recall: 0.5089 - val_AUC: 0.8036 - lr: 0.0025
Epoch 14/30
99/99 [==============================] - 4s 43ms/step - loss: 0.8253 - accuracy: 0.6006 - precision: 0.6429 - recall: 0.4987 - AUC: 0.7997 - val_loss: 0.9054 - val_accuracy: 0.6018 - val_precision: 0.6530 - val_recall: 0.5051 - val_AUC: 0.7897 - lr: 0.0022
Epoch 15/30
99/99 [==============================] - 4s 39ms/step - loss: 0.8179 - accuracy: 0.6137 - precision: 0.6610 - recall: 0.4936 - AUC: 0.8051 - val_loss: 0.9055 - val_accuracy: 0.6107 - val_precision: 0.6782 - val_recall: 0.4987 - val_AUC: 0.8033 - lr: 0.0020
Epoch 16/30
99/99 [==============================] - 3s 34ms/step - loss: 0.8001 - accuracy: 0.6204 - precision: 0.6698 - recall: 0.5232 - AUC: 0.8132 - val_loss: 0.8789 - val_accuracy: 0.6094 - val_precision: 0.6613 - val_recall: 0.5191 - val_AUC: 0.8075 - lr: 0.0018
Epoch 17/30
99/99 [==============================] - 3s 28ms/step - loss: 0.7980 - accuracy: 0.6191 - precision: 0.6645 - recall: 0.5191 - AUC: 0.8146 - val_loss: 0.8986 - val_accuracy: 0.5954 - val_precision: 0.6639 - val_recall: 0.5051 - val_AUC: 0.7972 - lr: 0.0016
time: 57.9 s (started: 2023-03-25 02:19:05 +00:00)


# Evaluate the model results and put into a dict
results = model8.evaluate(X_test, y_test)
results = dict(zip(model8.metrics_names,results))

print(results)
plottingScores(H)

53/53 [==============================] - 0s 7ms/step - loss: 0.7808 - accuracy: 0.6423 - precision: 0.6856 - recall: 0.5235 - AUC: 0.8277
{'loss': 0.780842661857605, 'accuracy': 0.6423054337501526, 'precision': 0.685603141784668, 'recall': 0.5234699845314026, 'AUC': 0.8276516795158386}

time: 3.04 s (started: 2023-03-25 02:20:03 +00:00)


y_pred = model8.predict(X_test)
# y_pred = np.array([0 if proba < 0.5 else 1 for proba in y_pred_proba])

53/53 [==============================] - 1s 6ms/step
time: 1.07 s (started: 2023-03-25 02:20:06 +00:00)


# Classification Report
print("Classification Report:\n",classification_report(y_test.argmax(axis=1), y_pred.argmax(axis=1)))

# Confusion Matrix
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test.argmax(axis=1), y_pred.argmax(axis=1))
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1', '2']],  
                         columns = [i for i in ['0', '1', '2']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.67      0.77      0.72       603
           1       0.53      0.50      0.52       600
           2       0.73      0.66      0.69       480

    accuracy                           0.64      1683
   macro avg       0.65      0.64      0.64      1683
weighted avg       0.64      0.64      0.64      1683

Confusion Matrix Chart:

time: 497 ms (started: 2023-03-25 02:20:07 +00:00)


# Model comparison
precision = precision_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
recall = recall_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
f1 = f1_score(y_test.argmax(axis=1),y_pred.argmax(axis=1), average='macro')
AUC = results['AUC']

Train_Accuracy = model8.evaluate(X_train, y_train)
Test_Accuracy = model8.evaluate(X_test, y_test)

# base_1 = []
base_1.append(['Model-8: CNN-1 (Tuned)', Train_Accuracy[1], Test_Accuracy[1], precision, recall, f1, AUC])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

123/123 [==============================] - 1s 8ms/step - loss: 0.7703 - accuracy: 0.6490 - precision: 0.7065 - recall: 0.5290 - AUC: 0.8340
53/53 [==============================] - 0s 7ms/step - loss: 0.7808 - accuracy: 0.6423 - precision: 0.6856 - recall: 0.5235 - AUC: 0.8277
time: 2.05 s (started: 2023-03-25 02:20:08 +00:00)


# Sumarize the results of modeling in a dataframe; Datapoints used = 3000; Epochs = 10/20; ADJUSTED_IMAGE_SIZE = 128
model_comparison

time: 9.12 ms (started: 2023-03-25 02:20:10 +00:00)


# Bar graph for Model Vs. F1 Score
plt.figure(figsize=(20,10))
sns.barplot(data=model_comparison, x="Model", y="F1 Score")
plt.title("Model Comparison: Model Vs. F1 Score")
plt.show()

time: 316 ms (started: 2023-03-25 02:20:10 +00:00)


# Bar graph for Model Vs. Metrics
model_comparison[['Model', 'Precision', 'Recall', 'F1 Score', 'AUC']].plot(kind='bar', x = 'Model', rot = 0, sort_columns = 'Model', figsize=(20,10))
plt.title("Model Comparison: Model Vs. Metrics")
plt.xlabel("Model")
plt.ylabel("Metrics")

Text(0, 0.5, 'Metrics')

time: 461 ms (started: 2023-03-25 02:20:10 +00:00)


# File paths from the G Drive
trainImagesPath = "stage_2_train_images"
testImagesPath = "stage_2_test_images"

labelsPath = "stage_2_train_labels.csv"
classInfoPath = "stage_2_detailed_class_info.csv"

# Read the labels and classinfo
labels = pd.read_csv(labelsPath)
details = pd.read_csv(classInfoPath)

time: 75.4 ms (started: 2023-03-25 02:37:26 +00:00)


labels.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 30227 entries, 0 to 30226
Data columns (total 6 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   patientId  30227 non-null  object 
 1   x          9555 non-null   float64
 2   y          9555 non-null   float64
 3   width      9555 non-null   float64
 4   height     9555 non-null   float64
 5   Target     30227 non-null  int64  
dtypes: float64(4), int64(1), object(1)
memory usage: 1.4+ MB
time: 31.2 ms (started: 2023-03-25 02:37:27 +00:00)


labels.sample(10)

time: 23.7 ms (started: 2023-03-25 02:37:27 +00:00)


details.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 30227 entries, 0 to 30226
Data columns (total 2 columns):
 #   Column     Non-Null Count  Dtype 
---  ------     --------------  ----- 
 0   patientId  30227 non-null  object
 1   class      30227 non-null  object
dtypes: object(2)
memory usage: 472.4+ KB
time: 24.8 ms (started: 2023-03-25 02:37:27 +00:00)


details.sample(10)

time: 8.2 ms (started: 2023-03-25 02:37:27 +00:00)


# Concatenate the two datasets - 'labels' and 'details':
train_data = pd.concat([labels, details['class']], axis = 1)

train_data.head()

time: 17.4 ms (started: 2023-03-25 02:37:27 +00:00)


train_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 30227 entries, 0 to 30226
Data columns (total 7 columns):
 #   Column     Non-Null Count  Dtype  
---  ------     --------------  -----  
 0   patientId  30227 non-null  object 
 1   x          9555 non-null   float64
 2   y          9555 non-null   float64
 3   width      9555 non-null   float64
 4   height     9555 non-null   float64
 5   Target     30227 non-null  int64  
 6   class      30227 non-null  object 
dtypes: float64(4), int64(1), object(2)
memory usage: 1.6+ MB
time: 20.3 ms (started: 2023-03-25 02:37:28 +00:00)


# Use the 2000 datapoints for training and 2000 for testing
# Use full datasets later
traind = train_data[0:2000]
testd = train_data[2000:4000]

traind.fillna(0, inplace=True)
testd.fillna(0, inplace=True)

time: 4.7 ms (started: 2023-03-25 02:37:28 +00:00)


# Check the distribution of the Target variable
traind.Target.value_counts()

0    1143
1     857
Name: Target, dtype: int64

time: 3.84 ms (started: 2023-03-25 02:37:28 +00:00)


# Check the distribution of the Target variable
testd.Target.value_counts()

1    1065
0     935
Name: Target, dtype: int64

time: 5.05 ms (started: 2023-03-25 02:37:28 +00:00)


# Keras data generator class; Refer the below articles for details:
# https://stanford.edu/~shervine/blog/keras-how-to-generate-data-on-the-fly
# https://towardsdatascience.com/keras-data-generators-and-how-to-use-them-b69129ed779c

# Manage the memory by reducing the batch size; 8, 10, 16, 32, 64
BATCH_SIZE = 6
# Image size to be scaled; 128 × 128, 224 x 224, 256 × 256, and 512 × 512
IMAGE_SIZE = 224

# Actual Image size 
IMG_WIDTH = 1024
IMG_HEIGHT = 1024

class TrainGenerator(Sequence):

    def __init__(self,  _labels):       
        self.pids = _labels["patientId"].to_numpy()
        self.coords = _labels[["x", "y", "width", "height"]].to_numpy()
        self.coords = self.coords * IMAGE_SIZE / IMG_WIDTH        

    def __len__(self):
        return math.ceil(len(self.coords) / BATCH_SIZE)

    """
    The contrast of an image is enhanced when various shades in the image becomes more distinct.
    We can do so by darkening the shades of the darker pixels and vice versa. 
    This is equivalent to widening the range of pixel intensities. To have a good contrast, 
    the following histogram characteristics are desirable:

    1) the pixel intensities are uniformly distributed across the full range of values (each intensity value is equally probable), and
    2) the cumulative histogram is increasing linearly across the full intensity range.

    Histogram equalization modifies the distribution of pixel intensities to achieve these characteristics.
    """
    def __doHistogramEqualization(self,img):
        # Pre processing Histogram equalization
        histogram_array = np.bincount(img.flatten(), minlength=256)
        # Normalize
        num_pixels = np.sum(histogram_array)
        histogram_array = histogram_array/num_pixels
        # Normalized cumulative histogram
        chistogram_array = np.cumsum(histogram_array)
        """
        STEP 2: Pixel mapping lookup table
        """
        transform_map = np.floor(255 * chistogram_array).astype(np.uint8)
        """
        STEP 3: Transformation
        """
        img_list = list(img.flatten())

        # Transform pixel values to equalize
        eq_img_list = [transform_map[p] for p in img_list]

        # Reshape and write back into img_array
        img = np.reshape(np.asarray(eq_img_list), img.shape)

        return img

    def __getitem__(self, idx): # Get a batch
        batch_coords = self.coords[idx * BATCH_SIZE:(idx + 1) * BATCH_SIZE] # Image coords
        batch_pids = self.pids[idx * BATCH_SIZE:(idx + 1) * BATCH_SIZE] # Image pids    
        batch_images = np.zeros((len(batch_pids), IMAGE_SIZE, IMAGE_SIZE, 3), dtype=np.float32)
        batch_masks = np.zeros((len(batch_pids), IMAGE_SIZE, IMAGE_SIZE))
        for _indx, _pid in enumerate(batch_pids):
            _path = 'stage_2_train_images/'+'{}.dcm'.format(_pid)
            _imgData = dcm.read_file(_path)

            img = _imgData.pixel_array 
            # img = np.stack((img,)*3, axis=-1) # Expand grayscale image to contain 3 channels

            # Resize image
            resized_img = cv2.resize(img,(IMAGE_SIZE,IMAGE_SIZE), interpolation = cv2.INTER_AREA)
            
            resized_img = self.__doHistogramEqualization(resized_img)
        
            batch_images[_indx][:,:,0] = preprocess_input(np.array(resized_img[:,:], dtype=np.float32)) 
            batch_images[_indx][:,:,1] = preprocess_input(np.array(resized_img[:,:], dtype=np.float32)) 
            batch_images[_indx][:,:,2] = preprocess_input(np.array(resized_img[:,:], dtype=np.float32)) 
            x = int(batch_coords[_indx, 0])
            y = int(batch_coords[_indx, 1])
            width = int(batch_coords[_indx, 2])
            height = int(batch_coords[_indx, 3])
            
            batch_masks[_indx][y:y+height, x:x+width] = 1

        return batch_images, batch_masks

time: 5.74 ms (started: 2023-03-25 02:37:28 +00:00)


traind = TrainGenerator(traind)
testd = TrainGenerator(testd)

time: 4.66 ms (started: 2023-03-25 02:37:28 +00:00)


# Function to show the images with mask
def showMaskedImage(_imageSet, _maskSet, _index) :
    maskImage = _imageSet[_index]

    maskImage[:,:,0] = _maskSet[_index] * _imageSet[_index][:,:,0]
    maskImage[:,:,1] = _maskSet[_index] * _imageSet[_index][:,:,1]
    maskImage[:,:,2] = _maskSet[_index] * _imageSet[_index][:,:,2]

    plt.imshow(maskImage[:,:,0])

time: 1.61 ms (started: 2023-03-25 02:37:28 +00:00)


# Sample pre-processed image from the TrainGenerator class
imageSet0 = traind[1][0][1]
plt.imshow(imageSet0)

WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).

<matplotlib.image.AxesImage at 0x7f9492fcdbb0>

time: 500 ms (started: 2023-03-25 02:37:28 +00:00)


# Masks for the same
imageSet0 = traind[2][0]
maskSet0 = traind[2][1]    
showMaskedImage(imageSet0, maskSet0, 5)

time: 558 ms (started: 2023-03-25 02:37:28 +00:00)


# Function to calculate the Dice Coef.
def dice_coef(y_true, y_pred):
    y_true = tf.keras.layers.Flatten()(y_true)
    y_pred = tf.keras.layers.Flatten()(y_pred)
    intersection = tf.reduce_sum(y_true * y_pred)
    return (2. * intersection + tf.keras.backend.epsilon()) / (tf.reduce_sum(y_true) + tf.reduce_sum(y_pred) + tf.keras.backend.epsilon())

# Dice Loss
def dice_loss(y_true, y_pred):
    return 1.0 - dice_coef(y_true, y_pred)

# # IoU Coef; Use the Keras metrics for MeanIoU
# def iou_coef(y_true, y_pred, smooth=1):
#   intersection = K.sum(K.abs(y_true * y_pred), axis=[1,2,3])
#   union = K.sum(y_true,[1,2,3])+K.sum(y_pred,[1,2,3])-intersection
#   iou = K.mean((intersection + smooth) / (union + smooth), axis=0)
#   return iou

# F1 Score: Taken from old keras source code
def f1_score(y_true, y_pred): 
    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
    possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))
    predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))
    precision = true_positives / (predicted_positives + K.epsilon())
    recall = true_positives / (possible_positives + K.epsilon())
    f1_val = 2*(precision*recall)/(precision+recall+K.epsilon())
    return f1_val

time: 1.82 ms (started: 2023-03-25 02:37:29 +00:00)


# Function to plot the various metrics across the epochs
"""
@Description: This function plots our metrics for our models across epochs
@Inputs: The history of the fitted model
@Output: Plots for accuracy, precision, recall, AUC, and loss
"""
def plottingScores(hist):
    fig, ax = plt.subplots(1, 8, figsize=(30, 4))
    ax = ax.ravel()

    for i, met in enumerate(['accuracy', 'precision', 'recall', 'f1_score', 'AUC', 'loss', 'dice_coef', 'mean_iou']):
        ax[i].plot(hist.history[met])
        ax[i].plot(hist.history['val_' + met])
        ax[i].set_title('Model {}'.format(met))
        ax[i].set_xlabel('epochs')
        ax[i].set_ylabel(met)
        ax[i].legend(['train', 'val'])

time: 1.26 ms (started: 2023-03-25 02:37:29 +00:00)


# Metrics to evaluate the model
METRICS = ['accuracy', 
           tf.keras.metrics.Precision(name='precision'), 
           tf.keras.metrics.Recall(name='recall'), 
           f1_score,
           tf.keras.metrics.AUC(name='AUC'),
           dice_coef,
           tf.keras.metrics.MeanIoU(num_classes=2, name='mean_iou')]

time: 18.6 ms (started: 2023-03-25 02:37:29 +00:00)


# Define the callback, checkpoint and early_stopping
def exponential_decay(lr0, s):
    def exponential_decay_fn(epoch):
        return lr0 * 0.1 **(epoch / s)
    return exponential_decay_fn

exponential_decay_fn = exponential_decay(0.01, 20)
lr_scheduler = tf.keras.callbacks.LearningRateScheduler(exponential_decay_fn)
checkpoint_cb = tf.keras.callbacks.ModelCheckpoint("unet_model.h5", save_best_only=True)
early_stopping_cb = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True)

time: 1.13 ms (started: 2023-03-25 02:37:29 +00:00)


LR = 1e-4
ALPHA = 1.0

# Define the UNet with MobileNet
def create_unet_mobilenet(trainable=True):
    """Function to create UNet architecture with MobileNet.
        
    Arguments:
        trainable -- Flag to make layers trainable. Default value is 'True'.
    """
    # Get all layers with 'imagenet' weights
    model = MobileNet(input_shape=(IMAGE_SIZE, IMAGE_SIZE, 3), include_top=False, alpha=ALPHA, weights="imagenet") 
    # Top layer is last layer of the model
    
    # Make all layers trainable
    for layer in model.layers:
        layer.trainable = trainable

    # Add all the UNET layers here
    convLayer_112by112 = model.get_layer("conv_pw_1_relu").output
    convLayer_56by56 = model.get_layer("conv_pw_3_relu").output
    convLayer_28by28 = model.get_layer("conv_pw_5_relu").output
    convLayer_14by14 = model.get_layer("conv_pw_11_relu").output
    convLayer_7by7 = model.get_layer("conv_pw_13_relu").output
    # The last layer of mobilenet model is of dimensions (7x7x1024)

    # Start upsampling from 7x7 to 14x14 ...up to 224x224 to form UNET
    # concatinate with the original image layer of the same size from MobileNet
    x = Concatenate()([UpSampling2D()(convLayer_7by7), convLayer_14by14])
    x = Concatenate()([UpSampling2D()(x), convLayer_28by28])
    x = Concatenate()([UpSampling2D()(x), convLayer_56by56])
    x = Concatenate()([UpSampling2D()(x), convLayer_112by112])
    x = UpSampling2D(name="unet_last")(x) # upsample to 224x224

    # Add classification layer
    x = Conv2D(1, kernel_size=1, activation="sigmoid", name="masks")(x)
    x = Reshape((IMAGE_SIZE, IMAGE_SIZE))(x) 
    
    return Model(inputs=model.input, outputs=x)

time: 2.02 ms (started: 2023-03-25 02:37:29 +00:00)


# Build the model 
input_shape = (IMAGE_SIZE, IMAGE_SIZE, 3)
model9 = create_unet_mobilenet(input_shape)
model9.summary()

Model: "model_1"
__________________________________________________________________________________________________
 Layer (type)                   Output Shape         Param #     Connected to                     
==================================================================================================
 input_6 (InputLayer)           [(None, 224, 224, 3  0           []                               
                                )]                                                                
                                                                                                  
 conv1 (Conv2D)                 (None, 112, 112, 32  864         ['input_6[0][0]']                
                                )                                                                 
                                                                                                  
 conv1_bn (BatchNormalization)  (None, 112, 112, 32  128         ['conv1[0][0]']                  
                                )                                                                 
                                                                                                  
 conv1_relu (ReLU)              (None, 112, 112, 32  0           ['conv1_bn[0][0]']               
                                )                                                                 
                                                                                                  
 conv_dw_1 (DepthwiseConv2D)    (None, 112, 112, 32  288         ['conv1_relu[0][0]']             
                                )                                                                 
                                                                                                  
 conv_dw_1_bn (BatchNormalizati  (None, 112, 112, 32  128        ['conv_dw_1[0][0]']              
 on)                            )                                                                 
                                                                                                  
 conv_dw_1_relu (ReLU)          (None, 112, 112, 32  0           ['conv_dw_1_bn[0][0]']           
                                )                                                                 
                                                                                                  
 conv_pw_1 (Conv2D)             (None, 112, 112, 64  2048        ['conv_dw_1_relu[0][0]']         
                                )                                                                 
                                                                                                  
 conv_pw_1_bn (BatchNormalizati  (None, 112, 112, 64  256        ['conv_pw_1[0][0]']              
 on)                            )                                                                 
                                                                                                  
 conv_pw_1_relu (ReLU)          (None, 112, 112, 64  0           ['conv_pw_1_bn[0][0]']           
                                )                                                                 
                                                                                                  
 conv_pad_2 (ZeroPadding2D)     (None, 113, 113, 64  0           ['conv_pw_1_relu[0][0]']         
                                )                                                                 
                                                                                                  
 conv_dw_2 (DepthwiseConv2D)    (None, 56, 56, 64)   576         ['conv_pad_2[0][0]']             
                                                                                                  
 conv_dw_2_bn (BatchNormalizati  (None, 56, 56, 64)  256         ['conv_dw_2[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_dw_2_relu (ReLU)          (None, 56, 56, 64)   0           ['conv_dw_2_bn[0][0]']           
                                                                                                  
 conv_pw_2 (Conv2D)             (None, 56, 56, 128)  8192        ['conv_dw_2_relu[0][0]']         
                                                                                                  
 conv_pw_2_bn (BatchNormalizati  (None, 56, 56, 128)  512        ['conv_pw_2[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_pw_2_relu (ReLU)          (None, 56, 56, 128)  0           ['conv_pw_2_bn[0][0]']           
                                                                                                  
 conv_dw_3 (DepthwiseConv2D)    (None, 56, 56, 128)  1152        ['conv_pw_2_relu[0][0]']         
                                                                                                  
 conv_dw_3_bn (BatchNormalizati  (None, 56, 56, 128)  512        ['conv_dw_3[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_dw_3_relu (ReLU)          (None, 56, 56, 128)  0           ['conv_dw_3_bn[0][0]']           
                                                                                                  
 conv_pw_3 (Conv2D)             (None, 56, 56, 128)  16384       ['conv_dw_3_relu[0][0]']         
                                                                                                  
 conv_pw_3_bn (BatchNormalizati  (None, 56, 56, 128)  512        ['conv_pw_3[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_pw_3_relu (ReLU)          (None, 56, 56, 128)  0           ['conv_pw_3_bn[0][0]']           
                                                                                                  
 conv_pad_4 (ZeroPadding2D)     (None, 57, 57, 128)  0           ['conv_pw_3_relu[0][0]']         
                                                                                                  
 conv_dw_4 (DepthwiseConv2D)    (None, 28, 28, 128)  1152        ['conv_pad_4[0][0]']             
                                                                                                  
 conv_dw_4_bn (BatchNormalizati  (None, 28, 28, 128)  512        ['conv_dw_4[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_dw_4_relu (ReLU)          (None, 28, 28, 128)  0           ['conv_dw_4_bn[0][0]']           
                                                                                                  
 conv_pw_4 (Conv2D)             (None, 28, 28, 256)  32768       ['conv_dw_4_relu[0][0]']         
                                                                                                  
 conv_pw_4_bn (BatchNormalizati  (None, 28, 28, 256)  1024       ['conv_pw_4[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_pw_4_relu (ReLU)          (None, 28, 28, 256)  0           ['conv_pw_4_bn[0][0]']           
                                                                                                  
 conv_dw_5 (DepthwiseConv2D)    (None, 28, 28, 256)  2304        ['conv_pw_4_relu[0][0]']         
                                                                                                  
 conv_dw_5_bn (BatchNormalizati  (None, 28, 28, 256)  1024       ['conv_dw_5[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_dw_5_relu (ReLU)          (None, 28, 28, 256)  0           ['conv_dw_5_bn[0][0]']           
                                                                                                  
 conv_pw_5 (Conv2D)             (None, 28, 28, 256)  65536       ['conv_dw_5_relu[0][0]']         
                                                                                                  
 conv_pw_5_bn (BatchNormalizati  (None, 28, 28, 256)  1024       ['conv_pw_5[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_pw_5_relu (ReLU)          (None, 28, 28, 256)  0           ['conv_pw_5_bn[0][0]']           
                                                                                                  
 conv_pad_6 (ZeroPadding2D)     (None, 29, 29, 256)  0           ['conv_pw_5_relu[0][0]']         
                                                                                                  
 conv_dw_6 (DepthwiseConv2D)    (None, 14, 14, 256)  2304        ['conv_pad_6[0][0]']             
                                                                                                  
 conv_dw_6_bn (BatchNormalizati  (None, 14, 14, 256)  1024       ['conv_dw_6[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_dw_6_relu (ReLU)          (None, 14, 14, 256)  0           ['conv_dw_6_bn[0][0]']           
                                                                                                  
 conv_pw_6 (Conv2D)             (None, 14, 14, 512)  131072      ['conv_dw_6_relu[0][0]']         
                                                                                                  
 conv_pw_6_bn (BatchNormalizati  (None, 14, 14, 512)  2048       ['conv_pw_6[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_pw_6_relu (ReLU)          (None, 14, 14, 512)  0           ['conv_pw_6_bn[0][0]']           
                                                                                                  
 conv_dw_7 (DepthwiseConv2D)    (None, 14, 14, 512)  4608        ['conv_pw_6_relu[0][0]']         
                                                                                                  
 conv_dw_7_bn (BatchNormalizati  (None, 14, 14, 512)  2048       ['conv_dw_7[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_dw_7_relu (ReLU)          (None, 14, 14, 512)  0           ['conv_dw_7_bn[0][0]']           
                                                                                                  
 conv_pw_7 (Conv2D)             (None, 14, 14, 512)  262144      ['conv_dw_7_relu[0][0]']         
                                                                                                  
 conv_pw_7_bn (BatchNormalizati  (None, 14, 14, 512)  2048       ['conv_pw_7[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_pw_7_relu (ReLU)          (None, 14, 14, 512)  0           ['conv_pw_7_bn[0][0]']           
                                                                                                  
 conv_dw_8 (DepthwiseConv2D)    (None, 14, 14, 512)  4608        ['conv_pw_7_relu[0][0]']         
                                                                                                  
 conv_dw_8_bn (BatchNormalizati  (None, 14, 14, 512)  2048       ['conv_dw_8[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_dw_8_relu (ReLU)          (None, 14, 14, 512)  0           ['conv_dw_8_bn[0][0]']           
                                                                                                  
 conv_pw_8 (Conv2D)             (None, 14, 14, 512)  262144      ['conv_dw_8_relu[0][0]']         
                                                                                                  
 conv_pw_8_bn (BatchNormalizati  (None, 14, 14, 512)  2048       ['conv_pw_8[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_pw_8_relu (ReLU)          (None, 14, 14, 512)  0           ['conv_pw_8_bn[0][0]']           
                                                                                                  
 conv_dw_9 (DepthwiseConv2D)    (None, 14, 14, 512)  4608        ['conv_pw_8_relu[0][0]']         
                                                                                                  
 conv_dw_9_bn (BatchNormalizati  (None, 14, 14, 512)  2048       ['conv_dw_9[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_dw_9_relu (ReLU)          (None, 14, 14, 512)  0           ['conv_dw_9_bn[0][0]']           
                                                                                                  
 conv_pw_9 (Conv2D)             (None, 14, 14, 512)  262144      ['conv_dw_9_relu[0][0]']         
                                                                                                  
 conv_pw_9_bn (BatchNormalizati  (None, 14, 14, 512)  2048       ['conv_pw_9[0][0]']              
 on)                                                                                              
                                                                                                  
 conv_pw_9_relu (ReLU)          (None, 14, 14, 512)  0           ['conv_pw_9_bn[0][0]']           
                                                                                                  
 conv_dw_10 (DepthwiseConv2D)   (None, 14, 14, 512)  4608        ['conv_pw_9_relu[0][0]']         
                                                                                                  
 conv_dw_10_bn (BatchNormalizat  (None, 14, 14, 512)  2048       ['conv_dw_10[0][0]']             
 ion)                                                                                             
                                                                                                  
 conv_dw_10_relu (ReLU)         (None, 14, 14, 512)  0           ['conv_dw_10_bn[0][0]']          
                                                                                                  
 conv_pw_10 (Conv2D)            (None, 14, 14, 512)  262144      ['conv_dw_10_relu[0][0]']        
                                                                                                  
 conv_pw_10_bn (BatchNormalizat  (None, 14, 14, 512)  2048       ['conv_pw_10[0][0]']             
 ion)                                                                                             
                                                                                                  
 conv_pw_10_relu (ReLU)         (None, 14, 14, 512)  0           ['conv_pw_10_bn[0][0]']          
                                                                                                  
 conv_dw_11 (DepthwiseConv2D)   (None, 14, 14, 512)  4608        ['conv_pw_10_relu[0][0]']        
                                                                                                  
 conv_dw_11_bn (BatchNormalizat  (None, 14, 14, 512)  2048       ['conv_dw_11[0][0]']             
 ion)                                                                                             
                                                                                                  
 conv_dw_11_relu (ReLU)         (None, 14, 14, 512)  0           ['conv_dw_11_bn[0][0]']          
                                                                                                  
 conv_pw_11 (Conv2D)            (None, 14, 14, 512)  262144      ['conv_dw_11_relu[0][0]']        
                                                                                                  
 conv_pw_11_bn (BatchNormalizat  (None, 14, 14, 512)  2048       ['conv_pw_11[0][0]']             
 ion)                                                                                             
                                                                                                  
 conv_pw_11_relu (ReLU)         (None, 14, 14, 512)  0           ['conv_pw_11_bn[0][0]']          
                                                                                                  
 conv_pad_12 (ZeroPadding2D)    (None, 15, 15, 512)  0           ['conv_pw_11_relu[0][0]']        
                                                                                                  
 conv_dw_12 (DepthwiseConv2D)   (None, 7, 7, 512)    4608        ['conv_pad_12[0][0]']            
                                                                                                  
 conv_dw_12_bn (BatchNormalizat  (None, 7, 7, 512)   2048        ['conv_dw_12[0][0]']             
 ion)                                                                                             
                                                                                                  
 conv_dw_12_relu (ReLU)         (None, 7, 7, 512)    0           ['conv_dw_12_bn[0][0]']          
                                                                                                  
 conv_pw_12 (Conv2D)            (None, 7, 7, 1024)   524288      ['conv_dw_12_relu[0][0]']        
                                                                                                  
 conv_pw_12_bn (BatchNormalizat  (None, 7, 7, 1024)  4096        ['conv_pw_12[0][0]']             
 ion)                                                                                             
                                                                                                  
 conv_pw_12_relu (ReLU)         (None, 7, 7, 1024)   0           ['conv_pw_12_bn[0][0]']          
                                                                                                  
 conv_dw_13 (DepthwiseConv2D)   (None, 7, 7, 1024)   9216        ['conv_pw_12_relu[0][0]']        
                                                                                                  
 conv_dw_13_bn (BatchNormalizat  (None, 7, 7, 1024)  4096        ['conv_dw_13[0][0]']             
 ion)                                                                                             
                                                                                                  
 conv_dw_13_relu (ReLU)         (None, 7, 7, 1024)   0           ['conv_dw_13_bn[0][0]']          
                                                                                                  
 conv_pw_13 (Conv2D)            (None, 7, 7, 1024)   1048576     ['conv_dw_13_relu[0][0]']        
                                                                                                  
 conv_pw_13_bn (BatchNormalizat  (None, 7, 7, 1024)  4096        ['conv_pw_13[0][0]']             
 ion)                                                                                             
                                                                                                  
 conv_pw_13_relu (ReLU)         (None, 7, 7, 1024)   0           ['conv_pw_13_bn[0][0]']          
                                                                                                  
 up_sampling2d_4 (UpSampling2D)  (None, 14, 14, 1024  0          ['conv_pw_13_relu[0][0]']        
                                )                                                                 
                                                                                                  
 concatenate_4 (Concatenate)    (None, 14, 14, 1536  0           ['up_sampling2d_4[0][0]',        
                                )                                 'conv_pw_11_relu[0][0]']        
                                                                                                  
 up_sampling2d_5 (UpSampling2D)  (None, 28, 28, 1536  0          ['concatenate_4[0][0]']          
                                )                                                                 
                                                                                                  
 concatenate_5 (Concatenate)    (None, 28, 28, 1792  0           ['up_sampling2d_5[0][0]',        
                                )                                 'conv_pw_5_relu[0][0]']         
                                                                                                  
 up_sampling2d_6 (UpSampling2D)  (None, 56, 56, 1792  0          ['concatenate_5[0][0]']          
                                )                                                                 
                                                                                                  
 concatenate_6 (Concatenate)    (None, 56, 56, 1920  0           ['up_sampling2d_6[0][0]',        
                                )                                 'conv_pw_3_relu[0][0]']         
                                                                                                  
 up_sampling2d_7 (UpSampling2D)  (None, 112, 112, 19  0          ['concatenate_6[0][0]']          
                                20)                                                               
                                                                                                  
 concatenate_7 (Concatenate)    (None, 112, 112, 19  0           ['up_sampling2d_7[0][0]',        
                                84)                               'conv_pw_1_relu[0][0]']         
                                                                                                  
 unet_last (UpSampling2D)       (None, 224, 224, 19  0           ['concatenate_7[0][0]']          
                                84)                                                               
                                                                                                  
 masks (Conv2D)                 (None, 224, 224, 1)  1985        ['unet_last[0][0]']              
                                                                                                  
 reshape_1 (Reshape)            (None, 224, 224)     0           ['masks[0][0]']                  
                                                                                                  
==================================================================================================
Total params: 3,230,849
Trainable params: 3,208,961
Non-trainable params: 21,888
__________________________________________________________________________________________________
time: 1.04 s (started: 2023-03-25 02:37:29 +00:00)


model9.compile(optimizer="adam", loss = dice_loss, metrics=METRICS)

time: 38.3 ms (started: 2023-03-25 02:37:30 +00:00)


# Fit the model
train_steps = len(traind)//BATCH_SIZE
valid_steps = len(testd)//BATCH_SIZE

if len(traind) % BATCH_SIZE != 0:
    train_steps += 1
if len(testd) % BATCH_SIZE != 0:
    valid_steps += 1


H = model9.fit(traind, epochs=20,
                        steps_per_epoch=train_steps,
                        validation_data=testd,
                        callbacks = [checkpoint_cb, early_stopping_cb, lr_scheduler],
                        use_multiprocessing=True,
                        workers=4,
                        validation_steps=valid_steps,                      
                        shuffle=True)
                        # class_weight = classWeight; Not supported for 3+ Dimensions

Epoch 1/20
56/56 [==============================] - 57s 683ms/step - loss: 0.8732 - accuracy: 0.0109 - precision: 0.0840 - recall: 0.3957 - f1_score: 0.1269 - AUC: 0.6380 - dice_coef: 0.1268 - mean_iou: 0.4824 - val_loss: 0.8929 - val_accuracy: 0.8940 - val_precision: 0.0000e+00 - val_recall: 0.0000e+00 - val_f1_score: 0.0000e+00 - val_AUC: 0.5000 - val_dice_coef: 0.1071 - val_mean_iou: 0.4902 - lr: 0.0100
Epoch 2/20
56/56 [==============================] - 36s 623ms/step - loss: 0.8298 - accuracy: 0.0022 - precision: 0.1244 - recall: 0.3371 - f1_score: 0.1701 - AUC: 0.6217 - dice_coef: 0.1702 - mean_iou: 0.4966 - val_loss: 0.9792 - val_accuracy: 0.6927 - val_precision: 0.0187 - val_recall: 0.0069 - val_f1_score: 0.0029 - val_AUC: 0.4988 - val_dice_coef: 0.0208 - val_mean_iou: 0.4911 - lr: 0.0089
Epoch 3/20
56/56 [==============================] - 37s 653ms/step - loss: 0.8739 - accuracy: 0.0051 - precision: 0.1156 - recall: 0.2182 - f1_score: 0.1260 - AUC: 0.5828 - dice_coef: 0.1261 - mean_iou: 0.5031 - val_loss: 0.8943 - val_accuracy: 0.8161 - val_precision: 0.1186 - val_recall: 0.0612 - val_f1_score: 0.0547 - val_AUC: 0.5305 - val_dice_coef: 0.1057 - val_mean_iou: 0.5014 - lr: 0.0079
Epoch 4/20
56/56 [==============================] - 39s 685ms/step - loss: 0.8354 - accuracy: 0.0114 - precision: 0.1594 - recall: 0.2247 - f1_score: 0.1646 - AUC: 0.5860 - dice_coef: 0.1646 - mean_iou: 0.5081 - val_loss: 0.8929 - val_accuracy: 0.7178 - val_precision: 0.0000e+00 - val_recall: 0.0000e+00 - val_f1_score: 0.0000e+00 - val_AUC: 0.4999 - val_dice_coef: 0.1071 - val_mean_iou: 0.4902 - lr: 0.0071
Epoch 5/20
56/56 [==============================] - 30s 517ms/step - loss: 0.8113 - accuracy: 0.0106 - precision: 0.1417 - recall: 0.4270 - f1_score: 0.1887 - AUC: 0.6663 - dice_coef: 0.1887 - mean_iou: 0.5044 - val_loss: 0.8685 - val_accuracy: 0.5513 - val_precision: 0.1300 - val_recall: 0.0543 - val_f1_score: 0.0436 - val_AUC: 0.5258 - val_dice_coef: 0.1315 - val_mean_iou: 0.5036 - lr: 0.0063
Epoch 6/20
56/56 [==============================] - 39s 682ms/step - loss: 0.7641 - accuracy: 0.0105 - precision: 0.1933 - recall: 0.4063 - f1_score: 0.2360 - AUC: 0.6708 - dice_coef: 0.2359 - mean_iou: 0.5309 - val_loss: 0.8548 - val_accuracy: 0.1991 - val_precision: 0.1201 - val_recall: 0.3189 - val_f1_score: 0.1452 - val_AUC: 0.6389 - val_dice_coef: 0.1452 - val_mean_iou: 0.5203 - lr: 0.0056
Epoch 7/20
56/56 [==============================] - 27s 452ms/step - loss: 0.7531 - accuracy: 0.0073 - precision: 0.1884 - recall: 0.5017 - f1_score: 0.2468 - AUC: 0.7139 - dice_coef: 0.2469 - mean_iou: 0.5345 - val_loss: 0.8854 - val_accuracy: 0.2879 - val_precision: 0.1299 - val_recall: 0.0716 - val_f1_score: 0.0431 - val_AUC: 0.5319 - val_dice_coef: 0.1146 - val_mean_iou: 0.5087 - lr: 0.0050
Epoch 8/20
56/56 [==============================] - 30s 524ms/step - loss: 0.7953 - accuracy: 0.0095 - precision: 0.1636 - recall: 0.4199 - f1_score: 0.2048 - AUC: 0.6750 - dice_coef: 0.2047 - mean_iou: 0.5209 - val_loss: 0.8273 - val_accuracy: 0.3727 - val_precision: 0.1411 - val_recall: 0.1686 - val_f1_score: 0.1042 - val_AUC: 0.5764 - val_dice_coef: 0.1727 - val_mean_iou: 0.5222 - lr: 0.0045
Epoch 9/20
56/56 [==============================] - 27s 474ms/step - loss: 0.7897 - accuracy: 0.0109 - precision: 0.1555 - recall: 0.5001 - f1_score: 0.2103 - AUC: 0.7059 - dice_coef: 0.2103 - mean_iou: 0.5185 - val_loss: 0.8831 - val_accuracy: 0.2842 - val_precision: 0.1246 - val_recall: 0.0711 - val_f1_score: 0.0449 - val_AUC: 0.5344 - val_dice_coef: 0.1169 - val_mean_iou: 0.5051 - lr: 0.0040
Epoch 10/20
56/56 [==============================] - 40s 680ms/step - loss: 0.7624 - accuracy: 0.0278 - precision: 0.1978 - recall: 0.4035 - f1_score: 0.2375 - AUC: 0.6711 - dice_coef: 0.2376 - mean_iou: 0.5339 - val_loss: 0.8758 - val_accuracy: 0.0964 - val_precision: 0.1221 - val_recall: 0.2387 - val_f1_score: 0.1241 - val_AUC: 0.6055 - val_dice_coef: 0.1242 - val_mean_iou: 0.5218 - lr: 0.0035
Epoch 11/20
56/56 [==============================] - 35s 617ms/step - loss: 0.7632 - accuracy: 0.0318 - precision: 0.1801 - recall: 0.5002 - f1_score: 0.2368 - AUC: 0.7150 - dice_coef: 0.2368 - mean_iou: 0.5346 - val_loss: 0.8393 - val_accuracy: 0.2346 - val_precision: 0.1346 - val_recall: 0.3228 - val_f1_score: 0.1607 - val_AUC: 0.6456 - val_dice_coef: 0.1607 - val_mean_iou: 0.5256 - lr: 0.0032
Epoch 12/20
56/56 [==============================] - 38s 654ms/step - loss: 0.7549 - accuracy: 0.0412 - precision: 0.2035 - recall: 0.4413 - f1_score: 0.2452 - AUC: 0.6875 - dice_coef: 0.2451 - mean_iou: 0.5358 - val_loss: 0.8200 - val_accuracy: 0.1155 - val_precision: 0.1340 - val_recall: 0.5016 - val_f1_score: 0.1801 - val_AUC: 0.7208 - val_dice_coef: 0.1800 - val_mean_iou: 0.5257 - lr: 0.0028
Epoch 13/20
56/56 [==============================] - 29s 506ms/step - loss: 0.7345 - accuracy: 0.0438 - precision: 0.2059 - recall: 0.5353 - f1_score: 0.2681 - AUC: 0.7352 - dice_coef: 0.2681 - mean_iou: 0.5477 - val_loss: 0.8146 - val_accuracy: 0.0635 - val_precision: 0.1363 - val_recall: 0.5111 - val_f1_score: 0.1854 - val_AUC: 0.7265 - val_dice_coef: 0.1854 - val_mean_iou: 0.5271 - lr: 0.0025
Epoch 14/20
56/56 [==============================] - 37s 648ms/step - loss: 0.7389 - accuracy: 0.0580 - precision: 0.2204 - recall: 0.4349 - f1_score: 0.2612 - AUC: 0.6947 - dice_coef: 0.2611 - mean_iou: 0.5528 - val_loss: 0.7666 - val_accuracy: 0.0776 - val_precision: 0.1497 - val_recall: 0.4408 - val_f1_score: 0.1798 - val_AUC: 0.6991 - val_dice_coef: 0.2334 - val_mean_iou: 0.5352 - lr: 0.0022
Epoch 15/20
56/56 [==============================] - 36s 624ms/step - loss: 0.7604 - accuracy: 0.0563 - precision: 0.2180 - recall: 0.3845 - f1_score: 0.2396 - AUC: 0.6695 - dice_coef: 0.2396 - mean_iou: 0.5472 - val_loss: 0.7895 - val_accuracy: 0.0625 - val_precision: 0.1881 - val_recall: 0.3511 - val_f1_score: 0.1933 - val_AUC: 0.6675 - val_dice_coef: 0.2105 - val_mean_iou: 0.5501 - lr: 0.0020
Epoch 16/20
56/56 [==============================] - 29s 489ms/step - loss: 0.7765 - accuracy: 0.0529 - precision: 0.2183 - recall: 0.3458 - f1_score: 0.2235 - AUC: 0.6542 - dice_coef: 0.2235 - mean_iou: 0.5462 - val_loss: 0.8054 - val_accuracy: 0.0340 - val_precision: 0.2556 - val_recall: 0.2018 - val_f1_score: 0.1593 - val_AUC: 0.6080 - val_dice_coef: 0.1946 - val_mean_iou: 0.5398 - lr: 0.0018
Epoch 17/20
56/56 [==============================] - 38s 653ms/step - loss: 0.7259 - accuracy: 0.0310 - precision: 0.2432 - recall: 0.4180 - f1_score: 0.2780 - AUC: 0.6844 - dice_coef: 0.2780 - mean_iou: 0.5507 - val_loss: 0.8044 - val_accuracy: 0.0121 - val_precision: 0.1222 - val_recall: 0.6929 - val_f1_score: 0.1957 - val_AUC: 0.8009 - val_dice_coef: 0.1956 - val_mean_iou: 0.5124 - lr: 0.0016
Epoch 18/20
56/56 [==============================] - 36s 622ms/step - loss: 0.7251 - accuracy: 0.0229 - precision: 0.2275 - recall: 0.4717 - f1_score: 0.2749 - AUC: 0.7073 - dice_coef: 0.2749 - mean_iou: 0.5530 - val_loss: 0.7934 - val_accuracy: 0.0149 - val_precision: 0.1339 - val_recall: 0.6806 - val_f1_score: 0.2065 - val_AUC: 0.8016 - val_dice_coef: 0.2066 - val_mean_iou: 0.5238 - lr: 0.0014
Epoch 19/20
56/56 [==============================] - 35s 620ms/step - loss: 0.7083 - accuracy: 0.0170 - precision: 0.2477 - recall: 0.4620 - f1_score: 0.2919 - AUC: 0.7081 - dice_coef: 0.2917 - mean_iou: 0.5597 - val_loss: 0.7964 - val_accuracy: 0.0054 - val_precision: 0.1316 - val_recall: 0.6715 - val_f1_score: 0.2036 - val_AUC: 0.8001 - val_dice_coef: 0.2036 - val_mean_iou: 0.5272 - lr: 0.0013
Epoch 20/20
56/56 [==============================] - 27s 449ms/step - loss: 0.6900 - accuracy: 0.0344 - precision: 0.2632 - recall: 0.4784 - f1_score: 0.3100 - AUC: 0.7185 - dice_coef: 0.3100 - mean_iou: 0.5675 - val_loss: 0.7834 - val_accuracy: 0.0230 - val_precision: 0.1412 - val_recall: 0.6766 - val_f1_score: 0.2166 - val_AUC: 0.8043 - val_dice_coef: 0.2166 - val_mean_iou: 0.5314 - lr: 0.0011
time: 13min 36s (started: 2023-03-25 02:37:30 +00:00)


# Evaluate and display results
results = model9.evaluate(testd) # Evaluate the model on test data
results = dict(zip(model9.metrics_names, results))

print(results)
plottingScores(H)

334/334 [==============================] - 745s 2s/step - loss: 0.7171 - accuracy: 0.0242 - precision: 0.2189 - recall: 0.6026 - f1_score: 0.2832 - AUC: 0.7675 - dice_coef: 0.2832 - mean_iou: 0.5569
{'loss': 0.7170584797859192, 'accuracy': 0.02421874925494194, 'precision': 0.21888083219528198, 'recall': 0.6025741100311279, 'f1_score': 0.2832229733467102, 'AUC': 0.7675365209579468, 'dice_coef': 0.2832002341747284, 'mean_iou': 0.5568951964378357}

time: 13min 24s (started: 2023-03-25 02:51:07 +00:00)


# Prepare the test data: Use random 20 images
testd1 = train_data[15000:15020]
testd1.fillna(0, inplace=True)

time: 2.58 ms (started: 2023-03-25 03:04:31 +00:00)


# Check target distrubution in test dataset (Equal distribution)
testd1.Target.value_counts()

0    10
1    10
Name: Target, dtype: int64

time: 6.57 ms (started: 2023-03-25 03:04:31 +00:00)


# Set the TrainGenerator for test data
testd1_datagen = TrainGenerator(testd1)

time: 3.58 ms (started: 2023-03-25 03:04:31 +00:00)


# Evaluate the model
test_steps = (len(testd1_datagen)//BATCH_SIZE)
if len(testd1_datagen) % BATCH_SIZE != 0:
    test_steps += 1

model9.evaluate(testd1_datagen)

4/4 [==============================] - 8s 3s/step - loss: 0.5486 - accuracy: 0.0263 - precision: 0.3695 - recall: 0.6399 - f1_score: 0.4095 - AUC: 0.7912 - dice_coef: 0.4097 - mean_iou: 0.6077

[0.5485594868659973,
 0.02633928507566452,
 0.36953845620155334,
 0.6398786306381226,
 0.4095156788825989,
 0.7912121415138245,
 0.40973907709121704,
 0.6076915264129639]

time: 13.3 s (started: 2023-03-25 03:04:31 +00:00)


# Prdict the test data that we have
pred_mask = model9.predict(testd1_datagen)

testd1 = testd1.reset_index()

4/4 [==============================] - 1s 161ms/step
time: 1.37 s (started: 2023-03-25 03:04:45 +00:00)


# Calculate the y_test, y_pred, tmpImages, tmpMask, originalMask

y_pred = []
y_test = []
imageList = []
predMaskTemp = []
IMAGE_HEIGHT = 224
IMAGE_WIDTH = 224

def getPredictions(testd1):
    masks = np.zeros((int(testd1.shape[0]), IMAGE_HEIGHT, IMAGE_WIDTH))

    for index, row in testd1.iterrows():
        patientId = row.patientId
        # print(patientId)

        classlabel = row["Target"]
        dcm_file = 'stage_2_train_images/'+'{}.dcm'.format(patientId)
        dcm_data = dcm.read_file(dcm_file)
        img = dcm_data.pixel_array
        resized_img = cv2.resize(img,(IMAGE_HEIGHT,IMAGE_WIDTH), interpolation = cv2.INTER_LINEAR)
        predMaskTemp.append(pred_mask[index])
        iou = (pred_mask[index] > 0.5) * 1.0
        y_pred.append((1 in iou) * 1)
        imageList.append(resized_img)
        y_test.append(classlabel)
        x_scale = IMAGE_HEIGHT / 1024
        y_scale = IMAGE_WIDTH / 1024

        if(classlabel == 1):
            x = int(np.round(row['x'] * x_scale))
            y = int(np.round(row['y'] * y_scale))
            w = int(np.round(row['width'] * x_scale))
            h = int(np.round(row['height'] * y_scale))
            masks[index][y:y+h, x:x+w] = 1
    tmpImages = np.array(imageList)
    tmpMask = np.array(predMaskTemp)
    originalMask = np.array(masks)
    return (y_test, y_pred, tmpImages, tmpMask, originalMask)

time: 3.89 ms (started: 2023-03-25 03:04:46 +00:00)


# Create predictions map
y_test, y_pred, imagelist, maskList, originalMask = getPredictions(testd1)

time: 275 ms (started: 2023-03-25 03:04:46 +00:00)


# Pick a random image
dcm_file = 'stage_2_train_images/'+'{}.dcm'.format('9358d1c5-ba61-4150-a233-41138208a3f9')
dcm_data = dcm.read_file(dcm_file)
img = dcm_data.pixel_array
plt.imshow(imagelist[12])

<matplotlib.image.AxesImage at 0x7f9493d4a280>

time: 347 ms (started: 2023-03-25 03:04:46 +00:00)


# Visualize the train and output data 

fig = plt.figure(figsize=(15, 15))

a = fig.add_subplot(1, 4, 1)
imgplot = plt.imshow(imagelist[1])
a.set_title('Original Images ',fontsize=20)

a = fig.add_subplot(1, 4, 2)
imgplot = plt.imshow(imagelist[12])

a = fig.add_subplot(1, 4, 3)
imgplot = plt.imshow(imagelist[13])

a = fig.add_subplot(1, 4, 4)
imgplot = plt.imshow(imagelist[15])

fig = plt.figure(figsize=(15, 15))
a = fig.add_subplot(1, 4, 1)

imgplot = plt.imshow(originalMask[1])
a.set_title('Oringial Mask (Truth) ',fontsize=20)
a.set_xlabel('Pneumonia {}:'.format(y_test[1]), fontsize=20)

a = fig.add_subplot(1, 4, 2)
imgplot = plt.imshow(originalMask[12])
a.set_xlabel('Pneumonia {}:'.format(y_test[12]), fontsize=20)

a = fig.add_subplot(1, 4, 3)
imgplot = plt.imshow(originalMask[13])
a.set_xlabel('Pneumonia {}:'.format(y_test[13]), fontsize=20)

a = fig.add_subplot(1, 4, 4)
imgplot = plt.imshow(originalMask[15])
a.set_xlabel('Pneumonia {}:'.format(y_test[15]), fontsize=20)

fig = plt.figure(figsize=(15, 15))
a = fig.add_subplot(1, 4, 1)
a.set_title('Predicted Mask  ',fontsize=20)
imgplot = plt.imshow(maskList[1])
a.set_xlabel('Pneumonia {}:'.format(y_pred[1]), fontsize=20)

a = fig.add_subplot(1, 4, 2)
imgplot = plt.imshow(maskList[12])
a.set_xlabel('Pneumonia {}:'.format(y_pred[12]), fontsize=20)

a = fig.add_subplot(1, 4, 3)
imgplot = plt.imshow(maskList[13])
a.set_xlabel('Pneumonia {}:'.format(y_pred[13]), fontsize=20)

a = fig.add_subplot(1, 4, 4)
imgplot = plt.imshow(maskList[15])
a.set_xlabel('Pneumonia {}:'.format(y_pred[15]), fontsize=20)

Text(0.5, 0, 'Pneumonia 1:')

time: 3.93 s (started: 2023-03-25 03:04:47 +00:00)


# Classification Report for test sample
print("Classification Report:\n",classification_report(y_test, y_pred))

# Confusion Matrix for test sample
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test, y_pred)
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1']],  
                         columns = [i for i in ['0', '1']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       1.00      0.60      0.75        10
           1       0.71      1.00      0.83        10

    accuracy                           0.80        20
   macro avg       0.86      0.80      0.79        20
weighted avg       0.86      0.80      0.79        20

Confusion Matrix Chart:

time: 423 ms (started: 2023-03-25 03:04:51 +00:00)


# Training history dataframe
hist = pd.DataFrame(H.history)

# Model comparison
Train_Accuracy = hist['accuracy'].mean()
# Test_Accuracy = model.evaluate(testd)

precision = results['precision']
recall = results['recall']
f1 = results['f1_score']
AUC = results['AUC']
Dice_Coef = results['dice_coef']
MeanIoU = results['mean_iou']

base_1 = []
base_1.append(['Model-9: UNet (MobileNet)', Train_Accuracy, results['accuracy'], precision, recall, f1, AUC, Dice_Coef, MeanIoU])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC', 'Dice Coef', 'MeanIoU'])
model_comparison.sort_values(by=['Recall','F1 Score'], inplace=True, ascending=False)

time: 9.82 ms (started: 2023-03-25 03:04:51 +00:00)


# Understand the various Metrics with epochs
hist

time: 34.5 ms (started: 2023-03-25 03:04:51 +00:00)


hist.describe()

time: 80 ms (started: 2023-03-25 03:04:51 +00:00)


# Function to create part of UNet architecture with ResNet50
def conv_block_simple(prevlayer, filters, prefix, strides=(1, 1)):
    """
    Adds convolutional layer followed by Batch Normalization and Activation layers.
        
    Arguments:
        prevlayer -- previous layer of the convolution block        
        filters -- number of filters for convolution
        prefix -- prefix for the layer name
        strides -- convolution stride. Default is 1x1.
    """
    conv = Conv2D(filters, (3, 3), padding="same", kernel_initializer="he_normal", strides=strides, name=prefix + "_conv")(prevlayer)
    conv = BatchNormalization(name=prefix + "_bn")(conv)
    conv = Activation('relu', name=prefix + "_activation")(conv)
    
    # Returns the built layers of the block.
    return conv

time: 3.86 ms (started: 2023-03-25 03:04:51 +00:00)


# Define the UNet with ResNet50
def create_unet_resnet50(trainable=True):
    """Function to create UNet architecture with ResNet50.
        
    Arguments:
        trainable -- Flag to make layers trainable. Default value is 'True'.
    """
    resnetLayers = ResNet50(weights='imagenet', input_shape=(IMAGE_SIZE, IMAGE_SIZE, 3), include_top=False) # Load pre-trained Resnet
    # Top layer is last layer of the model

    for layer in resnetLayers.layers:
        layer.trainable = trainable

    # Add all the UNet layers here
    convLayer_112by112 = resnetLayers.get_layer("conv1_relu").output
    convLayer_56by56 = resnetLayers.get_layer("conv2_block3_out").output # conv2_block3_2_relu
    convLayer_28by28 = resnetLayers.get_layer("conv3_block4_out").output # conv3_block4_2_relu
    convLayer_14by14 = resnetLayers.get_layer("conv4_block6_out").output # conv4_block6_2_relu
    convLayer_7by7 = resnetLayers.get_layer("conv5_block3_out").output # conv5_block3_2_relu
    # The last layer of resnet model(conv5_block3_out) is of dimensions (7x7x2048)
    # Start upsampling from 7x7 to 14x14 ...up to 224x224 to form UNet
    # concatinate with the original image layer of the same size from ResNet50
    up14by14 = Concatenate()([UpSampling2D()(convLayer_7by7), convLayer_14by14])
    upConvLayer_14by14 = conv_block_simple(up14by14, 256, "upConvLayer_14by14_1")
    upConvLayer_14by14 = conv_block_simple(upConvLayer_14by14, 256, "upConvLayer_14by14_2")
    
    up28by28 = Concatenate()([UpSampling2D()(upConvLayer_14by14), convLayer_28by28])
    upConvLayer_28by28 = conv_block_simple(up28by28, 256, "upConvLayer_28by28_1")
    upConvLayer_28by28 = conv_block_simple(upConvLayer_28by28, 256, "upConvLayer_28by28_2")
     
    up56by56 = Concatenate()([UpSampling2D()(upConvLayer_28by28), convLayer_56by56])
    upConvLayer_56by56 = conv_block_simple(up56by56, 256, "upConvLayer_56by56_1")
    upConvLayer_56by56 = conv_block_simple(upConvLayer_56by56, 256, "upConvLayer_56by56_2")    
    
    up112by112 = Concatenate()([UpSampling2D()(upConvLayer_56by56), convLayer_112by112])
    upConvLayer_112by112 = conv_block_simple(up112by112, 256, "upConvLayer_112by112_1")
    upConvLayer_112by112 = conv_block_simple(upConvLayer_112by112, 256, "upConvLayer_112by112_2")   
    
    up224by224 = UpSampling2D(name="unet_last")(upConvLayer_112by112) # upsample to 224x224
    upConvLayer_224by224 = conv_block_simple(up224by224, 256, "upConvLayer_224by224_1")
    upConvLayer_224by224 = conv_block_simple(upConvLayer_224by224, 256, "upConvLayer_224by224_2")
    # Add classification layer
    upConvLayer_224by224 = Conv2D(1, kernel_size=1, activation="sigmoid", name="masks")(upConvLayer_224by224)
    upConvLayer_224by224 = Reshape((IMAGE_SIZE, IMAGE_SIZE))(upConvLayer_224by224) 

    return Model(inputs=resnetLayers.input, outputs=upConvLayer_224by224)

time: 7.58 ms (started: 2023-03-25 03:04:51 +00:00)


# Build the model 
input_shape = (IMAGE_SIZE, IMAGE_SIZE, 3)
model10 = create_unet_resnet50(input_shape)
model10.summary()

Model: "model_2"
__________________________________________________________________________________________________
 Layer (type)                   Output Shape         Param #     Connected to                     
==================================================================================================
 input_7 (InputLayer)           [(None, 224, 224, 3  0           []                               
                                )]                                                                
                                                                                                  
 conv1_pad (ZeroPadding2D)      (None, 230, 230, 3)  0           ['input_7[0][0]']                
                                                                                                  
 conv1_conv (Conv2D)            (None, 112, 112, 64  9472        ['conv1_pad[0][0]']              
                                )                                                                 
                                                                                                  
 conv1_bn (BatchNormalization)  (None, 112, 112, 64  256         ['conv1_conv[0][0]']             
                                )                                                                 
                                                                                                  
 conv1_relu (Activation)        (None, 112, 112, 64  0           ['conv1_bn[0][0]']               
                                )                                                                 
                                                                                                  
 pool1_pad (ZeroPadding2D)      (None, 114, 114, 64  0           ['conv1_relu[0][0]']             
                                )                                                                 
                                                                                                  
 pool1_pool (MaxPooling2D)      (None, 56, 56, 64)   0           ['pool1_pad[0][0]']              
                                                                                                  
 conv2_block1_1_conv (Conv2D)   (None, 56, 56, 64)   4160        ['pool1_pool[0][0]']             
                                                                                                  
 conv2_block1_1_bn (BatchNormal  (None, 56, 56, 64)  256         ['conv2_block1_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block1_1_relu (Activatio  (None, 56, 56, 64)  0           ['conv2_block1_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv2_block1_2_conv (Conv2D)   (None, 56, 56, 64)   36928       ['conv2_block1_1_relu[0][0]']    
                                                                                                  
 conv2_block1_2_bn (BatchNormal  (None, 56, 56, 64)  256         ['conv2_block1_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block1_2_relu (Activatio  (None, 56, 56, 64)  0           ['conv2_block1_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv2_block1_0_conv (Conv2D)   (None, 56, 56, 256)  16640       ['pool1_pool[0][0]']             
                                                                                                  
 conv2_block1_3_conv (Conv2D)   (None, 56, 56, 256)  16640       ['conv2_block1_2_relu[0][0]']    
                                                                                                  
 conv2_block1_0_bn (BatchNormal  (None, 56, 56, 256)  1024       ['conv2_block1_0_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block1_3_bn (BatchNormal  (None, 56, 56, 256)  1024       ['conv2_block1_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block1_add (Add)         (None, 56, 56, 256)  0           ['conv2_block1_0_bn[0][0]',      
                                                                  'conv2_block1_3_bn[0][0]']      
                                                                                                  
 conv2_block1_out (Activation)  (None, 56, 56, 256)  0           ['conv2_block1_add[0][0]']       
                                                                                                  
 conv2_block2_1_conv (Conv2D)   (None, 56, 56, 64)   16448       ['conv2_block1_out[0][0]']       
                                                                                                  
 conv2_block2_1_bn (BatchNormal  (None, 56, 56, 64)  256         ['conv2_block2_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block2_1_relu (Activatio  (None, 56, 56, 64)  0           ['conv2_block2_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv2_block2_2_conv (Conv2D)   (None, 56, 56, 64)   36928       ['conv2_block2_1_relu[0][0]']    
                                                                                                  
 conv2_block2_2_bn (BatchNormal  (None, 56, 56, 64)  256         ['conv2_block2_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block2_2_relu (Activatio  (None, 56, 56, 64)  0           ['conv2_block2_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv2_block2_3_conv (Conv2D)   (None, 56, 56, 256)  16640       ['conv2_block2_2_relu[0][0]']    
                                                                                                  
 conv2_block2_3_bn (BatchNormal  (None, 56, 56, 256)  1024       ['conv2_block2_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block2_add (Add)         (None, 56, 56, 256)  0           ['conv2_block1_out[0][0]',       
                                                                  'conv2_block2_3_bn[0][0]']      
                                                                                                  
 conv2_block2_out (Activation)  (None, 56, 56, 256)  0           ['conv2_block2_add[0][0]']       
                                                                                                  
 conv2_block3_1_conv (Conv2D)   (None, 56, 56, 64)   16448       ['conv2_block2_out[0][0]']       
                                                                                                  
 conv2_block3_1_bn (BatchNormal  (None, 56, 56, 64)  256         ['conv2_block3_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block3_1_relu (Activatio  (None, 56, 56, 64)  0           ['conv2_block3_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv2_block3_2_conv (Conv2D)   (None, 56, 56, 64)   36928       ['conv2_block3_1_relu[0][0]']    
                                                                                                  
 conv2_block3_2_bn (BatchNormal  (None, 56, 56, 64)  256         ['conv2_block3_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block3_2_relu (Activatio  (None, 56, 56, 64)  0           ['conv2_block3_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv2_block3_3_conv (Conv2D)   (None, 56, 56, 256)  16640       ['conv2_block3_2_relu[0][0]']    
                                                                                                  
 conv2_block3_3_bn (BatchNormal  (None, 56, 56, 256)  1024       ['conv2_block3_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv2_block3_add (Add)         (None, 56, 56, 256)  0           ['conv2_block2_out[0][0]',       
                                                                  'conv2_block3_3_bn[0][0]']      
                                                                                                  
 conv2_block3_out (Activation)  (None, 56, 56, 256)  0           ['conv2_block3_add[0][0]']       
                                                                                                  
 conv3_block1_1_conv (Conv2D)   (None, 28, 28, 128)  32896       ['conv2_block3_out[0][0]']       
                                                                                                  
 conv3_block1_1_bn (BatchNormal  (None, 28, 28, 128)  512        ['conv3_block1_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block1_1_relu (Activatio  (None, 28, 28, 128)  0          ['conv3_block1_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv3_block1_2_conv (Conv2D)   (None, 28, 28, 128)  147584      ['conv3_block1_1_relu[0][0]']    
                                                                                                  
 conv3_block1_2_bn (BatchNormal  (None, 28, 28, 128)  512        ['conv3_block1_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block1_2_relu (Activatio  (None, 28, 28, 128)  0          ['conv3_block1_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv3_block1_0_conv (Conv2D)   (None, 28, 28, 512)  131584      ['conv2_block3_out[0][0]']       
                                                                                                  
 conv3_block1_3_conv (Conv2D)   (None, 28, 28, 512)  66048       ['conv3_block1_2_relu[0][0]']    
                                                                                                  
 conv3_block1_0_bn (BatchNormal  (None, 28, 28, 512)  2048       ['conv3_block1_0_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block1_3_bn (BatchNormal  (None, 28, 28, 512)  2048       ['conv3_block1_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block1_add (Add)         (None, 28, 28, 512)  0           ['conv3_block1_0_bn[0][0]',      
                                                                  'conv3_block1_3_bn[0][0]']      
                                                                                                  
 conv3_block1_out (Activation)  (None, 28, 28, 512)  0           ['conv3_block1_add[0][0]']       
                                                                                                  
 conv3_block2_1_conv (Conv2D)   (None, 28, 28, 128)  65664       ['conv3_block1_out[0][0]']       
                                                                                                  
 conv3_block2_1_bn (BatchNormal  (None, 28, 28, 128)  512        ['conv3_block2_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block2_1_relu (Activatio  (None, 28, 28, 128)  0          ['conv3_block2_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv3_block2_2_conv (Conv2D)   (None, 28, 28, 128)  147584      ['conv3_block2_1_relu[0][0]']    
                                                                                                  
 conv3_block2_2_bn (BatchNormal  (None, 28, 28, 128)  512        ['conv3_block2_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block2_2_relu (Activatio  (None, 28, 28, 128)  0          ['conv3_block2_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv3_block2_3_conv (Conv2D)   (None, 28, 28, 512)  66048       ['conv3_block2_2_relu[0][0]']    
                                                                                                  
 conv3_block2_3_bn (BatchNormal  (None, 28, 28, 512)  2048       ['conv3_block2_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block2_add (Add)         (None, 28, 28, 512)  0           ['conv3_block1_out[0][0]',       
                                                                  'conv3_block2_3_bn[0][0]']      
                                                                                                  
 conv3_block2_out (Activation)  (None, 28, 28, 512)  0           ['conv3_block2_add[0][0]']       
                                                                                                  
 conv3_block3_1_conv (Conv2D)   (None, 28, 28, 128)  65664       ['conv3_block2_out[0][0]']       
                                                                                                  
 conv3_block3_1_bn (BatchNormal  (None, 28, 28, 128)  512        ['conv3_block3_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block3_1_relu (Activatio  (None, 28, 28, 128)  0          ['conv3_block3_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv3_block3_2_conv (Conv2D)   (None, 28, 28, 128)  147584      ['conv3_block3_1_relu[0][0]']    
                                                                                                  
 conv3_block3_2_bn (BatchNormal  (None, 28, 28, 128)  512        ['conv3_block3_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block3_2_relu (Activatio  (None, 28, 28, 128)  0          ['conv3_block3_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv3_block3_3_conv (Conv2D)   (None, 28, 28, 512)  66048       ['conv3_block3_2_relu[0][0]']    
                                                                                                  
 conv3_block3_3_bn (BatchNormal  (None, 28, 28, 512)  2048       ['conv3_block3_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block3_add (Add)         (None, 28, 28, 512)  0           ['conv3_block2_out[0][0]',       
                                                                  'conv3_block3_3_bn[0][0]']      
                                                                                                  
 conv3_block3_out (Activation)  (None, 28, 28, 512)  0           ['conv3_block3_add[0][0]']       
                                                                                                  
 conv3_block4_1_conv (Conv2D)   (None, 28, 28, 128)  65664       ['conv3_block3_out[0][0]']       
                                                                                                  
 conv3_block4_1_bn (BatchNormal  (None, 28, 28, 128)  512        ['conv3_block4_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block4_1_relu (Activatio  (None, 28, 28, 128)  0          ['conv3_block4_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv3_block4_2_conv (Conv2D)   (None, 28, 28, 128)  147584      ['conv3_block4_1_relu[0][0]']    
                                                                                                  
 conv3_block4_2_bn (BatchNormal  (None, 28, 28, 128)  512        ['conv3_block4_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block4_2_relu (Activatio  (None, 28, 28, 128)  0          ['conv3_block4_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv3_block4_3_conv (Conv2D)   (None, 28, 28, 512)  66048       ['conv3_block4_2_relu[0][0]']    
                                                                                                  
 conv3_block4_3_bn (BatchNormal  (None, 28, 28, 512)  2048       ['conv3_block4_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv3_block4_add (Add)         (None, 28, 28, 512)  0           ['conv3_block3_out[0][0]',       
                                                                  'conv3_block4_3_bn[0][0]']      
                                                                                                  
 conv3_block4_out (Activation)  (None, 28, 28, 512)  0           ['conv3_block4_add[0][0]']       
                                                                                                  
 conv4_block1_1_conv (Conv2D)   (None, 14, 14, 256)  131328      ['conv3_block4_out[0][0]']       
                                                                                                  
 conv4_block1_1_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block1_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block1_1_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block1_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block1_2_conv (Conv2D)   (None, 14, 14, 256)  590080      ['conv4_block1_1_relu[0][0]']    
                                                                                                  
 conv4_block1_2_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block1_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block1_2_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block1_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block1_0_conv (Conv2D)   (None, 14, 14, 1024  525312      ['conv3_block4_out[0][0]']       
                                )                                                                 
                                                                                                  
 conv4_block1_3_conv (Conv2D)   (None, 14, 14, 1024  263168      ['conv4_block1_2_relu[0][0]']    
                                )                                                                 
                                                                                                  
 conv4_block1_0_bn (BatchNormal  (None, 14, 14, 1024  4096       ['conv4_block1_0_conv[0][0]']    
 ization)                       )                                                                 
                                                                                                  
 conv4_block1_3_bn (BatchNormal  (None, 14, 14, 1024  4096       ['conv4_block1_3_conv[0][0]']    
 ization)                       )                                                                 
                                                                                                  
 conv4_block1_add (Add)         (None, 14, 14, 1024  0           ['conv4_block1_0_bn[0][0]',      
                                )                                 'conv4_block1_3_bn[0][0]']      
                                                                                                  
 conv4_block1_out (Activation)  (None, 14, 14, 1024  0           ['conv4_block1_add[0][0]']       
                                )                                                                 
                                                                                                  
 conv4_block2_1_conv (Conv2D)   (None, 14, 14, 256)  262400      ['conv4_block1_out[0][0]']       
                                                                                                  
 conv4_block2_1_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block2_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block2_1_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block2_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block2_2_conv (Conv2D)   (None, 14, 14, 256)  590080      ['conv4_block2_1_relu[0][0]']    
                                                                                                  
 conv4_block2_2_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block2_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block2_2_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block2_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block2_3_conv (Conv2D)   (None, 14, 14, 1024  263168      ['conv4_block2_2_relu[0][0]']    
                                )                                                                 
                                                                                                  
 conv4_block2_3_bn (BatchNormal  (None, 14, 14, 1024  4096       ['conv4_block2_3_conv[0][0]']    
 ization)                       )                                                                 
                                                                                                  
 conv4_block2_add (Add)         (None, 14, 14, 1024  0           ['conv4_block1_out[0][0]',       
                                )                                 'conv4_block2_3_bn[0][0]']      
                                                                                                  
 conv4_block2_out (Activation)  (None, 14, 14, 1024  0           ['conv4_block2_add[0][0]']       
                                )                                                                 
                                                                                                  
 conv4_block3_1_conv (Conv2D)   (None, 14, 14, 256)  262400      ['conv4_block2_out[0][0]']       
                                                                                                  
 conv4_block3_1_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block3_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block3_1_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block3_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block3_2_conv (Conv2D)   (None, 14, 14, 256)  590080      ['conv4_block3_1_relu[0][0]']    
                                                                                                  
 conv4_block3_2_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block3_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block3_2_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block3_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block3_3_conv (Conv2D)   (None, 14, 14, 1024  263168      ['conv4_block3_2_relu[0][0]']    
                                )                                                                 
                                                                                                  
 conv4_block3_3_bn (BatchNormal  (None, 14, 14, 1024  4096       ['conv4_block3_3_conv[0][0]']    
 ization)                       )                                                                 
                                                                                                  
 conv4_block3_add (Add)         (None, 14, 14, 1024  0           ['conv4_block2_out[0][0]',       
                                )                                 'conv4_block3_3_bn[0][0]']      
                                                                                                  
 conv4_block3_out (Activation)  (None, 14, 14, 1024  0           ['conv4_block3_add[0][0]']       
                                )                                                                 
                                                                                                  
 conv4_block4_1_conv (Conv2D)   (None, 14, 14, 256)  262400      ['conv4_block3_out[0][0]']       
                                                                                                  
 conv4_block4_1_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block4_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block4_1_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block4_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block4_2_conv (Conv2D)   (None, 14, 14, 256)  590080      ['conv4_block4_1_relu[0][0]']    
                                                                                                  
 conv4_block4_2_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block4_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block4_2_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block4_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block4_3_conv (Conv2D)   (None, 14, 14, 1024  263168      ['conv4_block4_2_relu[0][0]']    
                                )                                                                 
                                                                                                  
 conv4_block4_3_bn (BatchNormal  (None, 14, 14, 1024  4096       ['conv4_block4_3_conv[0][0]']    
 ization)                       )                                                                 
                                                                                                  
 conv4_block4_add (Add)         (None, 14, 14, 1024  0           ['conv4_block3_out[0][0]',       
                                )                                 'conv4_block4_3_bn[0][0]']      
                                                                                                  
 conv4_block4_out (Activation)  (None, 14, 14, 1024  0           ['conv4_block4_add[0][0]']       
                                )                                                                 
                                                                                                  
 conv4_block5_1_conv (Conv2D)   (None, 14, 14, 256)  262400      ['conv4_block4_out[0][0]']       
                                                                                                  
 conv4_block5_1_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block5_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block5_1_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block5_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block5_2_conv (Conv2D)   (None, 14, 14, 256)  590080      ['conv4_block5_1_relu[0][0]']    
                                                                                                  
 conv4_block5_2_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block5_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block5_2_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block5_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block5_3_conv (Conv2D)   (None, 14, 14, 1024  263168      ['conv4_block5_2_relu[0][0]']    
                                )                                                                 
                                                                                                  
 conv4_block5_3_bn (BatchNormal  (None, 14, 14, 1024  4096       ['conv4_block5_3_conv[0][0]']    
 ization)                       )                                                                 
                                                                                                  
 conv4_block5_add (Add)         (None, 14, 14, 1024  0           ['conv4_block4_out[0][0]',       
                                )                                 'conv4_block5_3_bn[0][0]']      
                                                                                                  
 conv4_block5_out (Activation)  (None, 14, 14, 1024  0           ['conv4_block5_add[0][0]']       
                                )                                                                 
                                                                                                  
 conv4_block6_1_conv (Conv2D)   (None, 14, 14, 256)  262400      ['conv4_block5_out[0][0]']       
                                                                                                  
 conv4_block6_1_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block6_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block6_1_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block6_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block6_2_conv (Conv2D)   (None, 14, 14, 256)  590080      ['conv4_block6_1_relu[0][0]']    
                                                                                                  
 conv4_block6_2_bn (BatchNormal  (None, 14, 14, 256)  1024       ['conv4_block6_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv4_block6_2_relu (Activatio  (None, 14, 14, 256)  0          ['conv4_block6_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv4_block6_3_conv (Conv2D)   (None, 14, 14, 1024  263168      ['conv4_block6_2_relu[0][0]']    
                                )                                                                 
                                                                                                  
 conv4_block6_3_bn (BatchNormal  (None, 14, 14, 1024  4096       ['conv4_block6_3_conv[0][0]']    
 ization)                       )                                                                 
                                                                                                  
 conv4_block6_add (Add)         (None, 14, 14, 1024  0           ['conv4_block5_out[0][0]',       
                                )                                 'conv4_block6_3_bn[0][0]']      
                                                                                                  
 conv4_block6_out (Activation)  (None, 14, 14, 1024  0           ['conv4_block6_add[0][0]']       
                                )                                                                 
                                                                                                  
 conv5_block1_1_conv (Conv2D)   (None, 7, 7, 512)    524800      ['conv4_block6_out[0][0]']       
                                                                                                  
 conv5_block1_1_bn (BatchNormal  (None, 7, 7, 512)   2048        ['conv5_block1_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block1_1_relu (Activatio  (None, 7, 7, 512)   0           ['conv5_block1_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv5_block1_2_conv (Conv2D)   (None, 7, 7, 512)    2359808     ['conv5_block1_1_relu[0][0]']    
                                                                                                  
 conv5_block1_2_bn (BatchNormal  (None, 7, 7, 512)   2048        ['conv5_block1_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block1_2_relu (Activatio  (None, 7, 7, 512)   0           ['conv5_block1_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv5_block1_0_conv (Conv2D)   (None, 7, 7, 2048)   2099200     ['conv4_block6_out[0][0]']       
                                                                                                  
 conv5_block1_3_conv (Conv2D)   (None, 7, 7, 2048)   1050624     ['conv5_block1_2_relu[0][0]']    
                                                                                                  
 conv5_block1_0_bn (BatchNormal  (None, 7, 7, 2048)  8192        ['conv5_block1_0_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block1_3_bn (BatchNormal  (None, 7, 7, 2048)  8192        ['conv5_block1_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block1_add (Add)         (None, 7, 7, 2048)   0           ['conv5_block1_0_bn[0][0]',      
                                                                  'conv5_block1_3_bn[0][0]']      
                                                                                                  
 conv5_block1_out (Activation)  (None, 7, 7, 2048)   0           ['conv5_block1_add[0][0]']       
                                                                                                  
 conv5_block2_1_conv (Conv2D)   (None, 7, 7, 512)    1049088     ['conv5_block1_out[0][0]']       
                                                                                                  
 conv5_block2_1_bn (BatchNormal  (None, 7, 7, 512)   2048        ['conv5_block2_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block2_1_relu (Activatio  (None, 7, 7, 512)   0           ['conv5_block2_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv5_block2_2_conv (Conv2D)   (None, 7, 7, 512)    2359808     ['conv5_block2_1_relu[0][0]']    
                                                                                                  
 conv5_block2_2_bn (BatchNormal  (None, 7, 7, 512)   2048        ['conv5_block2_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block2_2_relu (Activatio  (None, 7, 7, 512)   0           ['conv5_block2_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv5_block2_3_conv (Conv2D)   (None, 7, 7, 2048)   1050624     ['conv5_block2_2_relu[0][0]']    
                                                                                                  
 conv5_block2_3_bn (BatchNormal  (None, 7, 7, 2048)  8192        ['conv5_block2_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block2_add (Add)         (None, 7, 7, 2048)   0           ['conv5_block1_out[0][0]',       
                                                                  'conv5_block2_3_bn[0][0]']      
                                                                                                  
 conv5_block2_out (Activation)  (None, 7, 7, 2048)   0           ['conv5_block2_add[0][0]']       
                                                                                                  
 conv5_block3_1_conv (Conv2D)   (None, 7, 7, 512)    1049088     ['conv5_block2_out[0][0]']       
                                                                                                  
 conv5_block3_1_bn (BatchNormal  (None, 7, 7, 512)   2048        ['conv5_block3_1_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block3_1_relu (Activatio  (None, 7, 7, 512)   0           ['conv5_block3_1_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv5_block3_2_conv (Conv2D)   (None, 7, 7, 512)    2359808     ['conv5_block3_1_relu[0][0]']    
                                                                                                  
 conv5_block3_2_bn (BatchNormal  (None, 7, 7, 512)   2048        ['conv5_block3_2_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block3_2_relu (Activatio  (None, 7, 7, 512)   0           ['conv5_block3_2_bn[0][0]']      
 n)                                                                                               
                                                                                                  
 conv5_block3_3_conv (Conv2D)   (None, 7, 7, 2048)   1050624     ['conv5_block3_2_relu[0][0]']    
                                                                                                  
 conv5_block3_3_bn (BatchNormal  (None, 7, 7, 2048)  8192        ['conv5_block3_3_conv[0][0]']    
 ization)                                                                                         
                                                                                                  
 conv5_block3_add (Add)         (None, 7, 7, 2048)   0           ['conv5_block2_out[0][0]',       
                                                                  'conv5_block3_3_bn[0][0]']      
                                                                                                  
 conv5_block3_out (Activation)  (None, 7, 7, 2048)   0           ['conv5_block3_add[0][0]']       
                                                                                                  
 up_sampling2d_8 (UpSampling2D)  (None, 14, 14, 2048  0          ['conv5_block3_out[0][0]']       
                                )                                                                 
                                                                                                  
 concatenate_8 (Concatenate)    (None, 14, 14, 3072  0           ['up_sampling2d_8[0][0]',        
                                )                                 'conv4_block6_out[0][0]']       
                                                                                                  
 upConvLayer_14by14_1_conv (Con  (None, 14, 14, 256)  7078144    ['concatenate_8[0][0]']          
 v2D)                                                                                             
                                                                                                  
 upConvLayer_14by14_1_bn (Batch  (None, 14, 14, 256)  1024       ['upConvLayer_14by14_1_conv[0][0]
 Normalization)                                                  ']                               
                                                                                                  
 upConvLayer_14by14_1_activatio  (None, 14, 14, 256)  0          ['upConvLayer_14by14_1_bn[0][0]']
 n (Activation)                                                                                   
                                                                                                  
 upConvLayer_14by14_2_conv (Con  (None, 14, 14, 256)  590080     ['upConvLayer_14by14_1_activation
 v2D)                                                            [0][0]']                         
                                                                                                  
 upConvLayer_14by14_2_bn (Batch  (None, 14, 14, 256)  1024       ['upConvLayer_14by14_2_conv[0][0]
 Normalization)                                                  ']                               
                                                                                                  
 upConvLayer_14by14_2_activatio  (None, 14, 14, 256)  0          ['upConvLayer_14by14_2_bn[0][0]']
 n (Activation)                                                                                   
                                                                                                  
 up_sampling2d_9 (UpSampling2D)  (None, 28, 28, 256)  0          ['upConvLayer_14by14_2_activation
                                                                 [0][0]']                         
                                                                                                  
 concatenate_9 (Concatenate)    (None, 28, 28, 768)  0           ['up_sampling2d_9[0][0]',        
                                                                  'conv3_block4_out[0][0]']       
                                                                                                  
 upConvLayer_28by28_1_conv (Con  (None, 28, 28, 256)  1769728    ['concatenate_9[0][0]']          
 v2D)                                                                                             
                                                                                                  
 upConvLayer_28by28_1_bn (Batch  (None, 28, 28, 256)  1024       ['upConvLayer_28by28_1_conv[0][0]
 Normalization)                                                  ']                               
                                                                                                  
 upConvLayer_28by28_1_activatio  (None, 28, 28, 256)  0          ['upConvLayer_28by28_1_bn[0][0]']
 n (Activation)                                                                                   
                                                                                                  
 upConvLayer_28by28_2_conv (Con  (None, 28, 28, 256)  590080     ['upConvLayer_28by28_1_activation
 v2D)                                                            [0][0]']                         
                                                                                                  
 upConvLayer_28by28_2_bn (Batch  (None, 28, 28, 256)  1024       ['upConvLayer_28by28_2_conv[0][0]
 Normalization)                                                  ']                               
                                                                                                  
 upConvLayer_28by28_2_activatio  (None, 28, 28, 256)  0          ['upConvLayer_28by28_2_bn[0][0]']
 n (Activation)                                                                                   
                                                                                                  
 up_sampling2d_10 (UpSampling2D  (None, 56, 56, 256)  0          ['upConvLayer_28by28_2_activation
 )                                                               [0][0]']                         
                                                                                                  
 concatenate_10 (Concatenate)   (None, 56, 56, 512)  0           ['up_sampling2d_10[0][0]',       
                                                                  'conv2_block3_out[0][0]']       
                                                                                                  
 upConvLayer_56by56_1_conv (Con  (None, 56, 56, 256)  1179904    ['concatenate_10[0][0]']         
 v2D)                                                                                             
                                                                                                  
 upConvLayer_56by56_1_bn (Batch  (None, 56, 56, 256)  1024       ['upConvLayer_56by56_1_conv[0][0]
 Normalization)                                                  ']                               
                                                                                                  
 upConvLayer_56by56_1_activatio  (None, 56, 56, 256)  0          ['upConvLayer_56by56_1_bn[0][0]']
 n (Activation)                                                                                   
                                                                                                  
 upConvLayer_56by56_2_conv (Con  (None, 56, 56, 256)  590080     ['upConvLayer_56by56_1_activation
 v2D)                                                            [0][0]']                         
                                                                                                  
 upConvLayer_56by56_2_bn (Batch  (None, 56, 56, 256)  1024       ['upConvLayer_56by56_2_conv[0][0]
 Normalization)                                                  ']                               
                                                                                                  
 upConvLayer_56by56_2_activatio  (None, 56, 56, 256)  0          ['upConvLayer_56by56_2_bn[0][0]']
 n (Activation)                                                                                   
                                                                                                  
 up_sampling2d_11 (UpSampling2D  (None, 112, 112, 25  0          ['upConvLayer_56by56_2_activation
 )                              6)                               [0][0]']                         
                                                                                                  
 concatenate_11 (Concatenate)   (None, 112, 112, 32  0           ['up_sampling2d_11[0][0]',       
                                0)                                'conv1_relu[0][0]']             
                                                                                                  
 upConvLayer_112by112_1_conv (C  (None, 112, 112, 25  737536     ['concatenate_11[0][0]']         
 onv2D)                         6)                                                                
                                                                                                  
 upConvLayer_112by112_1_bn (Bat  (None, 112, 112, 25  1024       ['upConvLayer_112by112_1_conv[0][
 chNormalization)               6)                               0]']                             
                                                                                                  
 upConvLayer_112by112_1_activat  (None, 112, 112, 25  0          ['upConvLayer_112by112_1_bn[0][0]
 ion (Activation)               6)                               ']                               
                                                                                                  
 upConvLayer_112by112_2_conv (C  (None, 112, 112, 25  590080     ['upConvLayer_112by112_1_activati
 onv2D)                         6)                               on[0][0]']                       
                                                                                                  
 upConvLayer_112by112_2_bn (Bat  (None, 112, 112, 25  1024       ['upConvLayer_112by112_2_conv[0][
 chNormalization)               6)                               0]']                             
                                                                                                  
 upConvLayer_112by112_2_activat  (None, 112, 112, 25  0          ['upConvLayer_112by112_2_bn[0][0]
 ion (Activation)               6)                               ']                               
                                                                                                  
 unet_last (UpSampling2D)       (None, 224, 224, 25  0           ['upConvLayer_112by112_2_activati
                                6)                               on[0][0]']                       
                                                                                                  
 upConvLayer_224by224_1_conv (C  (None, 224, 224, 25  590080     ['unet_last[0][0]']              
 onv2D)                         6)                                                                
                                                                                                  
 upConvLayer_224by224_1_bn (Bat  (None, 224, 224, 25  1024       ['upConvLayer_224by224_1_conv[0][
 chNormalization)               6)                               0]']                             
                                                                                                  
 upConvLayer_224by224_1_activat  (None, 224, 224, 25  0          ['upConvLayer_224by224_1_bn[0][0]
 ion (Activation)               6)                               ']                               
                                                                                                  
 upConvLayer_224by224_2_conv (C  (None, 224, 224, 25  590080     ['upConvLayer_224by224_1_activati
 onv2D)                         6)                               on[0][0]']                       
                                                                                                  
 upConvLayer_224by224_2_bn (Bat  (None, 224, 224, 25  1024       ['upConvLayer_224by224_2_conv[0][
 chNormalization)               6)                               0]']                             
                                                                                                  
 upConvLayer_224by224_2_activat  (None, 224, 224, 25  0          ['upConvLayer_224by224_2_bn[0][0]
 ion (Activation)               6)                               ']                               
                                                                                                  
 masks (Conv2D)                 (None, 224, 224, 1)  257         ['upConvLayer_224by224_2_activati
                                                                 on[0][0]']                       
                                                                                                  
 reshape_2 (Reshape)            (None, 224, 224)     0           ['masks[0][0]']                  
                                                                                                  
==================================================================================================
Total params: 37,904,001
Trainable params: 37,845,761
Non-trainable params: 58,240
__________________________________________________________________________________________________
time: 3.46 s (started: 2023-03-25 03:04:51 +00:00)


# Compile the model
model10.compile(optimizer="adam", loss = dice_loss, metrics=METRICS)

time: 23.6 ms (started: 2023-03-25 03:04:55 +00:00)


# Fit the model
train_steps = len(traind)//BATCH_SIZE
valid_steps = len(testd)//BATCH_SIZE

if len(traind) % BATCH_SIZE != 0:
    train_steps += 1
if len(testd) % BATCH_SIZE != 0:
    valid_steps += 1

H = model10.fit(traind, epochs=20,
                        steps_per_epoch=train_steps,
                        validation_data=testd,
                        callbacks = [checkpoint_cb, early_stopping_cb, lr_scheduler],
                        use_multiprocessing=True,
                        workers=4,
                        validation_steps=valid_steps,                      
                        shuffle=True)
                        # class_weight = classWeight; Not supported for 3+ Dimensions

Epoch 1/20
56/56 [==============================] - 114s 1s/step - loss: 0.8742 - accuracy: 0.0252 - precision: 0.0780 - recall: 0.5227 - f1_score: 0.1261 - AUC: 0.6416 - dice_coef: 0.1258 - mean_iou: 0.5079 - val_loss: 1.0000 - val_accuracy: 0.8940 - val_precision: 0.0000e+00 - val_recall: 0.0000e+00 - val_f1_score: 0.0000e+00 - val_AUC: 0.4999 - val_dice_coef: 3.7877e-10 - val_mean_iou: 0.4902 - lr: 0.0100
Epoch 2/20
56/56 [==============================] - 59s 1s/step - loss: 0.8054 - accuracy: 0.0053 - precision: 0.1374 - recall: 0.3655 - f1_score: 0.1948 - AUC: 0.6484 - dice_coef: 0.1946 - mean_iou: 0.5197 - val_loss: 0.9599 - val_accuracy: 0.8940 - val_precision: 0.0207 - val_recall: 0.9981 - val_f1_score: 0.0401 - val_AUC: 0.5279 - val_dice_coef: 0.0401 - val_mean_iou: 0.0392 - lr: 0.0089
Epoch 3/20
56/56 [==============================] - 64s 1s/step - loss: 0.7742 - accuracy: 0.0042 - precision: 0.2204 - recall: 0.3119 - f1_score: 0.2295 - AUC: 0.6468 - dice_coef: 0.2294 - mean_iou: 0.5435 - val_loss: 0.9961 - val_accuracy: 0.8940 - val_precision: 0.0025 - val_recall: 0.0219 - val_f1_score: 0.0039 - val_AUC: 0.4247 - val_dice_coef: 0.0039 - val_mean_iou: 0.4069 - lr: 0.0079
Epoch 4/20
56/56 [==============================] - 57s 1s/step - loss: 0.7792 - accuracy: 0.0051 - precision: 0.1774 - recall: 0.4458 - f1_score: 0.2209 - AUC: 0.6893 - dice_coef: 0.2208 - mean_iou: 0.5295 - val_loss: 0.9965 - val_accuracy: 0.8940 - val_precision: 0.0024 - val_recall: 0.0168 - val_f1_score: 0.0035 - val_AUC: 0.4379 - val_dice_coef: 0.0035 - val_mean_iou: 0.4225 - lr: 0.0071
Epoch 5/20
56/56 [==============================] - 55s 962ms/step - loss: 0.7502 - accuracy: 0.0038 - precision: 0.2245 - recall: 0.3605 - f1_score: 0.2550 - AUC: 0.6657 - dice_coef: 0.2550 - mean_iou: 0.5461 - val_loss: 0.9586 - val_accuracy: 0.8895 - val_precision: 0.0217 - val_recall: 0.6706 - val_f1_score: 0.0414 - val_AUC: 0.5342 - val_dice_coef: 0.0414 - val_mean_iou: 0.2099 - lr: 0.0063
Epoch 6/20
56/56 [==============================] - 57s 1s/step - loss: 0.8125 - accuracy: 0.0035 - precision: 0.1495 - recall: 0.3478 - f1_score: 0.1876 - AUC: 0.6537 - dice_coef: 0.1875 - mean_iou: 0.5270 - val_loss: 0.9590 - val_accuracy: 0.8611 - val_precision: 0.0212 - val_recall: 1.0000 - val_f1_score: 0.0410 - val_AUC: 0.5401 - val_dice_coef: 0.0410 - val_mean_iou: 0.0533 - lr: 0.0056
Epoch 7/20
56/56 [==============================] - 53s 934ms/step - loss: 0.7273 - accuracy: 0.0030 - precision: 0.2478 - recall: 0.3970 - f1_score: 0.2778 - AUC: 0.6863 - dice_coef: 0.2777 - mean_iou: 0.5561 - val_loss: 0.9473 - val_accuracy: 1.0629e-04 - val_precision: 0.0276 - val_recall: 0.9550 - val_f1_score: 0.0527 - val_AUC: 0.6475 - val_dice_coef: 0.0527 - val_mean_iou: 0.1986 - lr: 0.0050
Epoch 8/20
56/56 [==============================] - 58s 1s/step - loss: 0.7579 - accuracy: 0.0026 - precision: 0.2140 - recall: 0.3852 - f1_score: 0.2421 - AUC: 0.6756 - dice_coef: 0.2421 - mean_iou: 0.5415 - val_loss: 0.9461 - val_accuracy: 0.0220 - val_precision: 0.0281 - val_recall: 0.7405 - val_f1_score: 0.0539 - val_AUC: 0.6173 - val_dice_coef: 0.0539 - val_mean_iou: 0.2722 - lr: 0.0045
Epoch 9/20
56/56 [==============================] - 54s 938ms/step - loss: 0.7543 - accuracy: 0.0056 - precision: 0.2023 - recall: 0.4203 - f1_score: 0.2487 - AUC: 0.6836 - dice_coef: 0.2486 - mean_iou: 0.5410 - val_loss: 0.8624 - val_accuracy: 0.0086 - val_precision: 0.1575 - val_recall: 0.1430 - val_f1_score: 0.1377 - val_AUC: 0.5749 - val_dice_coef: 0.1376 - val_mean_iou: 0.5164 - lr: 0.0040
Epoch 10/20
56/56 [==============================] - 58s 1s/step - loss: 0.7340 - accuracy: 0.0065 - precision: 0.2332 - recall: 0.4433 - f1_score: 0.2661 - AUC: 0.7025 - dice_coef: 0.2660 - mean_iou: 0.5536 - val_loss: 0.8737 - val_accuracy: 0.0308 - val_precision: 0.0754 - val_recall: 0.6882 - val_f1_score: 0.1263 - val_AUC: 0.7656 - val_dice_coef: 0.1263 - val_mean_iou: 0.4633 - lr: 0.0035
Epoch 11/20
56/56 [==============================] - 58s 1s/step - loss: 0.7247 - accuracy: 0.0424 - precision: 0.2264 - recall: 0.4730 - f1_score: 0.2754 - AUC: 0.7111 - dice_coef: 0.2753 - mean_iou: 0.5518 - val_loss: 0.9137 - val_accuracy: 0.0043 - val_precision: 0.0495 - val_recall: 0.6396 - val_f1_score: 0.0863 - val_AUC: 0.7017 - val_dice_coef: 0.0863 - val_mean_iou: 0.4157 - lr: 0.0032
Epoch 12/20
56/56 [==============================] - 53s 918ms/step - loss: 0.7317 - accuracy: 0.0081 - precision: 0.2267 - recall: 0.4132 - f1_score: 0.2684 - AUC: 0.6864 - dice_coef: 0.2683 - mean_iou: 0.5508 - val_loss: 0.8552 - val_accuracy: 7.9719e-04 - val_precision: 0.0860 - val_recall: 0.8079 - val_f1_score: 0.1448 - val_AUC: 0.8256 - val_dice_coef: 0.1448 - val_mean_iou: 0.4742 - lr: 0.0028
Epoch 13/20
56/56 [==============================] - 57s 1s/step - loss: 0.7539 - accuracy: 0.0045 - precision: 0.2122 - recall: 0.4394 - f1_score: 0.2461 - AUC: 0.7035 - dice_coef: 0.2461 - mean_iou: 0.5512 - val_loss: 0.8904 - val_accuracy: 0.0076 - val_precision: 0.0643 - val_recall: 0.7129 - val_f1_score: 0.1096 - val_AUC: 0.7585 - val_dice_coef: 0.1096 - val_mean_iou: 0.4463 - lr: 0.0025
Epoch 14/20
56/56 [==============================] - 58s 1s/step - loss: 0.6884 - accuracy: 0.0046 - precision: 0.2731 - recall: 0.4867 - f1_score: 0.3153 - AUC: 0.7268 - dice_coef: 0.3152 - mean_iou: 0.5693 - val_loss: 0.8347 - val_accuracy: 0.0058 - val_precision: 0.2611 - val_recall: 0.2151 - val_f1_score: 0.1651 - val_AUC: 0.6114 - val_dice_coef: 0.1653 - val_mean_iou: 0.5425 - lr: 0.0022
Epoch 15/20
56/56 [==============================] - 63s 1s/step - loss: 0.7501 - accuracy: 0.0047 - precision: 0.2074 - recall: 0.4481 - f1_score: 0.2500 - AUC: 0.7043 - dice_coef: 0.2499 - mean_iou: 0.5506 - val_loss: 0.7620 - val_accuracy: 0.0029 - val_precision: 0.1903 - val_recall: 0.4254 - val_f1_score: 0.2382 - val_AUC: 0.7082 - val_dice_coef: 0.2380 - val_mean_iou: 0.5560 - lr: 0.0020
Epoch 16/20
56/56 [==============================] - 75s 1s/step - loss: 0.7087 - accuracy: 0.0063 - precision: 0.2581 - recall: 0.4304 - f1_score: 0.2914 - AUC: 0.7000 - dice_coef: 0.2913 - mean_iou: 0.5608 - val_loss: 0.7417 - val_accuracy: 0.0090 - val_precision: 0.2004 - val_recall: 0.5270 - val_f1_score: 0.2584 - val_AUC: 0.7548 - val_dice_coef: 0.2583 - val_mean_iou: 0.5672 - lr: 0.0018
Epoch 17/20
56/56 [==============================] - 58s 1s/step - loss: 0.7117 - accuracy: 0.0085 - precision: 0.2437 - recall: 0.4316 - f1_score: 0.2883 - AUC: 0.6998 - dice_coef: 0.2883 - mean_iou: 0.5615 - val_loss: 0.7853 - val_accuracy: 0.0033 - val_precision: 0.1620 - val_recall: 0.4696 - val_f1_score: 0.2148 - val_AUC: 0.7233 - val_dice_coef: 0.2147 - val_mean_iou: 0.5469 - lr: 0.0016
Epoch 18/20
56/56 [==============================] - 58s 1s/step - loss: 0.7335 - accuracy: 0.0071 - precision: 0.2330 - recall: 0.4178 - f1_score: 0.2666 - AUC: 0.6946 - dice_coef: 0.2665 - mean_iou: 0.5559 - val_loss: 0.7898 - val_accuracy: 0.0079 - val_precision: 0.2085 - val_recall: 0.3451 - val_f1_score: 0.2102 - val_AUC: 0.6694 - val_dice_coef: 0.2102 - val_mean_iou: 0.5558 - lr: 0.0014
Epoch 19/20
56/56 [==============================] - 57s 1s/step - loss: 0.7157 - accuracy: 0.0058 - precision: 0.2388 - recall: 0.4661 - f1_score: 0.2843 - AUC: 0.7191 - dice_coef: 0.2843 - mean_iou: 0.5632 - val_loss: 0.7557 - val_accuracy: 0.0077 - val_precision: 0.1696 - val_recall: 0.6276 - val_f1_score: 0.2444 - val_AUC: 0.7903 - val_dice_coef: 0.2443 - val_mean_iou: 0.5552 - lr: 0.0013
Epoch 20/20
56/56 [==============================] - 52s 901ms/step - loss: 0.7099 - accuracy: 0.0065 - precision: 0.2508 - recall: 0.4282 - f1_score: 0.2901 - AUC: 0.6979 - dice_coef: 0.2901 - mean_iou: 0.5626 - val_loss: 0.8277 - val_accuracy: 0.0021 - val_precision: 0.1098 - val_recall: 0.7121 - val_f1_score: 0.1723 - val_AUC: 0.8067 - val_dice_coef: 0.1723 - val_mean_iou: 0.5106 - lr: 0.0011
time: 26min 25s (started: 2023-03-25 03:04:55 +00:00)


# Evaluate and display results
results = model10.evaluate(testd) # Evaluate the model on test data
results = dict(zip(model10.metrics_names, results))

print(results)
plottingScores(H)

334/334 [==============================] - 83s 247ms/step - loss: 0.7608 - accuracy: 0.0028 - precision: 0.1735 - recall: 0.6327 - f1_score: 0.2394 - AUC: 0.7686 - dice_coef: 0.2394 - mean_iou: 0.5336
{'loss': 0.7608000636100769, 'accuracy': 0.0028013393748551607, 'precision': 0.17347915470600128, 'recall': 0.6326538324356079, 'f1_score': 0.23942965269088745, 'AUC': 0.7686296105384827, 'dice_coef': 0.23938621580600739, 'mean_iou': 0.533566415309906}

time: 1min 26s (started: 2023-03-25 03:31:21 +00:00)


# Prepare the test data: Use random 20 images
testd1 = train_data[15000:15020]
testd1.fillna(0, inplace=True)

time: 26.4 ms (started: 2023-03-25 03:32:47 +00:00)


# Check target distrubution in test dataset (Equal distribution)
testd1.Target.value_counts()

0    10
1    10
Name: Target, dtype: int64

time: 24.3 ms (started: 2023-03-25 03:32:47 +00:00)


# Set the TrainGenerator for test data
testd1_datagen = TrainGenerator(testd1)

time: 19.1 ms (started: 2023-03-25 03:32:47 +00:00)


# Evaluate the model
test_steps = (len(testd1_datagen)//BATCH_SIZE)
if len(testd1_datagen) % BATCH_SIZE != 0:
    test_steps += 1

model10.evaluate(testd1_datagen)

4/4 [==============================] - 1s 185ms/step - loss: 0.5885 - accuracy: 0.0000e+00 - precision: 0.3285 - recall: 0.6594 - f1_score: 0.3430 - AUC: 0.7972 - dice_coef: 0.3429 - mean_iou: 0.5942

[0.5884883403778076,
 0.0,
 0.3285060524940491,
 0.6593831777572632,
 0.3429979681968689,
 0.7972458004951477,
 0.34292837977409363,
 0.5941668748855591]

time: 1.27 s (started: 2023-03-25 03:32:47 +00:00)


# Predict the test data that we have
pred_mask = model10.predict(testd1_datagen)

testd1 = testd1.reset_index()

4/4 [==============================] - 2s 215ms/step
time: 2.32 s (started: 2023-03-25 03:32:48 +00:00)


# Calculate the y_test, y_pred, tmpImages, tmpMask, originalMask

y_pred = []
y_test = []
imageList = []
predMaskTemp = []
IMAGE_HEIGHT = 224
IMAGE_WIDTH = 224

def getPredictions(testd1):
    masks = np.zeros((int(testd1.shape[0]), IMAGE_HEIGHT, IMAGE_WIDTH))

    for index, row in testd1.iterrows():
        patientId = row.patientId
        # print(patientId)

        classlabel = row["Target"]
        dcm_file = 'stage_2_train_images/'+'{}.dcm'.format(patientId)
        dcm_data = dcm.read_file(dcm_file)
        img = dcm_data.pixel_array
        resized_img = cv2.resize(img,(IMAGE_HEIGHT,IMAGE_WIDTH), interpolation = cv2.INTER_LINEAR)
        predMaskTemp.append(pred_mask[index])
        iou = (pred_mask[index] > 0.5) * 1.0
        y_pred.append((1 in iou) * 1)
        imageList.append(resized_img)
        y_test.append(classlabel)
        x_scale = IMAGE_HEIGHT / 1024
        y_scale = IMAGE_WIDTH / 1024

        if(classlabel == 1):
            x = int(np.round(row['x'] * x_scale))
            y = int(np.round(row['y'] * y_scale))
            w = int(np.round(row['width'] * x_scale))
            h = int(np.round(row['height'] * y_scale))
            masks[index][y:y+h, x:x+w] = 1
    tmpImages = np.array(imageList)
    tmpMask = np.array(predMaskTemp)
    originalMask = np.array(masks)
    return (y_test, y_pred, tmpImages, tmpMask, originalMask)

time: 4.21 ms (started: 2023-03-25 03:32:50 +00:00)


# Create the predictions map
y_test, y_pred, imagelist, maskList, originalMask = getPredictions(testd1)

time: 215 ms (started: 2023-03-25 03:32:50 +00:00)


# Pick a random image
dcm_file = 'stage_2_train_images/'+'{}.dcm'.format('9358d1c5-ba61-4150-a233-41138208a3f9')
dcm_data = dcm.read_file(dcm_file)
img = dcm_data.pixel_array
plt.imshow(imagelist[12])

<matplotlib.image.AxesImage at 0x7f9356b31b80>

time: 359 ms (started: 2023-03-25 03:32:51 +00:00)


# Visualize the train and output data 

fig = plt.figure(figsize=(15, 15))

a = fig.add_subplot(1, 4, 1)
imgplot = plt.imshow(imagelist[1])
a.set_title('Original Images ',fontsize=20)

a = fig.add_subplot(1, 4, 2)
imgplot = plt.imshow(imagelist[12])

a = fig.add_subplot(1, 4, 3)
imgplot = plt.imshow(imagelist[13])

a = fig.add_subplot(1, 4, 4)
imgplot = plt.imshow(imagelist[15])

fig = plt.figure(figsize=(15, 15))
a = fig.add_subplot(1, 4, 1)

imgplot = plt.imshow(originalMask[1])
a.set_title('Oringial Mask (Truth) ',fontsize=20)
a.set_xlabel('Pneumonia {}:'.format(y_test[1]), fontsize=20)

a = fig.add_subplot(1, 4, 2)
imgplot = plt.imshow(originalMask[12])
a.set_xlabel('Pneumonia {}:'.format(y_test[12]), fontsize=20)

a = fig.add_subplot(1, 4, 3)
imgplot = plt.imshow(originalMask[13])
a.set_xlabel('Pneumonia {}:'.format(y_test[13]), fontsize=20)

a = fig.add_subplot(1, 4, 4)
imgplot = plt.imshow(originalMask[15])
a.set_xlabel('Pneumonia {}:'.format(y_test[15]), fontsize=20)

fig = plt.figure(figsize=(15, 15))
a = fig.add_subplot(1, 4, 1)
a.set_title('Predicted Mask  ',fontsize=20)
imgplot = plt.imshow(maskList[1])
a.set_xlabel('Pneumonia {}:'.format(y_pred[1]), fontsize=20)

a = fig.add_subplot(1, 4, 2)
imgplot = plt.imshow(maskList[12])
a.set_xlabel('Pneumonia {}:'.format(y_pred[12]), fontsize=20)

a = fig.add_subplot(1, 4, 3)
imgplot = plt.imshow(maskList[13])
a.set_xlabel('Pneumonia {}:'.format(y_pred[13]), fontsize=20)

a = fig.add_subplot(1, 4, 4)
imgplot = plt.imshow(maskList[15])
a.set_xlabel('Pneumonia {}:'.format(y_pred[15]), fontsize=20)

Text(0.5, 0, 'Pneumonia 1:')

time: 3.31 s (started: 2023-03-25 03:32:51 +00:00)


# Classification Report for test sample
print("Classification Report:\n",classification_report(y_test, y_pred))

# Confusion Matrix for test sample
print("Confusion Matrix Chart:")
cm = confusion_matrix(y_test, y_pred)
df_cm = pd.DataFrame(cm, index = [i for i in ['0', '1']],  
                         columns = [i for i in ['0', '1']])
plt.figure(figsize = (12,10))
sns.heatmap(df_cm, annot=True, fmt='g')
plt.title('Confusion Matrix')
plt.ylabel('Actual Label')
plt.xlabel('Predicted Label')
plt.show()

Classification Report:
               precision    recall  f1-score   support

           0       0.83      0.50      0.62        10
           1       0.64      0.90      0.75        10

    accuracy                           0.70        20
   macro avg       0.74      0.70      0.69        20
weighted avg       0.74      0.70      0.69        20

Confusion Matrix Chart:

time: 288 ms (started: 2023-03-25 03:32:54 +00:00)


# Training history dataframe
hist = pd.DataFrame(H.history)

# Model comparison
Train_Accuracy = hist['accuracy'].mean()
# Test_Accuracy = model.evaluate(testd)

precision = results['precision']
recall = results['recall']
f1 = results['f1_score']
AUC = results['AUC']
Dice_Coef = results['dice_coef']
MeanIoU = results['mean_iou']

# base_1 = []
base_1.append(['Model-10: UNet (ResNet50)', Train_Accuracy, results['accuracy'], precision, recall, f1, AUC, Dice_Coef, MeanIoU])
model_comparison = pd.DataFrame(base_1,columns=['Model','Train Accuracy','Test Accuracy','Precision','Recall','F1 Score', 'AUC', 'Dice Coef', 'MeanIoU'])
model_comparison.sort_values(by=['Dice Coef','MeanIoU'], inplace=True, ascending=False)

time: 17.1 ms (started: 2023-03-25 03:32:55 +00:00)


# Sumarize the results of modeling in a dataframe; Datapoints used = 4000; Epochs = 10/20; ADJUSTED_IMAGE_SIZE = 224
model_comparison

time: 13 ms (started: 2023-03-25 03:32:55 +00:00)


# Bar graph for Model Vs. F1 Score
plt.figure(figsize=(20,10))
sns.barplot(data=model_comparison, x="Model", y="MeanIoU")
plt.title("Model Comparison: Model Vs. MeanIoU")
plt.show()

time: 329 ms (started: 2023-03-25 03:32:55 +00:00)


# Bar graph for Model Vs. Metrics
model_comparison[['Model', 'Precision', 'Recall', 'F1 Score', 'AUC', 'Dice Coef', 'MeanIoU']].plot(kind='bar', x = 'Model', rot = 0, sort_columns = 'Model', figsize=(20,10))
plt.title("Model Comparison: Model Vs. Metrics")
plt.xlabel("Model")
plt.ylabel("Metrics")

Text(0, 0.5, 'Metrics')

time: 573 ms (started: 2023-03-25 03:32:55 +00:00)


# Save the Classification model to a file in the current working directory
Pkl_Filename = "Pickle_Classification_EfficientNetV2B3_Model.pkl"  

with open(Pkl_Filename, 'wb') as file:  
    pickle.dump(model7, file)

# # Load the DL Model back from the file
# with open(Pkl_Filename, 'rb') as file:  
#     Pickle_Classification_EfficientNetV2B3_Model = pickle.load(file)
    
# Pickle_Classification_EfficientNetV2B3_Model

Keras weights file (<HDF5 file "variables.h5" (mode r+)>) saving:
...layers
......dense
.........vars
............0
............1
......functional
.........layers
............activation
...............vars
............activation_1
...............vars
............activation_10
...............vars
............activation_11
...............vars
............activation_12
...............vars
............activation_13
...............vars
............activation_14
...............vars
............activation_15
...............vars
............activation_16
...............vars
............activation_17
...............vars
............activation_18
...............vars
............activation_19
...............vars
............activation_2
...............vars
............activation_20
...............vars
............activation_21
...............vars
............activation_22
...............vars
............activation_23
...............vars
............activation_24
...............vars
............activation_25
...............vars
............activation_26
...............vars
............activation_27
...............vars
............activation_28
...............vars
............activation_29
...............vars
............activation_3
...............vars
............activation_30
...............vars
............activation_31
...............vars
............activation_32
...............vars
............activation_33
...............vars
............activation_34
...............vars
............activation_35
...............vars
............activation_36
...............vars
............activation_37
...............vars
............activation_38
...............vars
............activation_39
...............vars
............activation_4
...............vars
............activation_40
...............vars
............activation_41
...............vars
............activation_42
...............vars
............activation_43
...............vars
............activation_44
...............vars
............activation_45
...............vars
............activation_46
...............vars
............activation_47
...............vars
............activation_48
...............vars
............activation_49
...............vars
............activation_5
...............vars
............activation_50
...............vars
............activation_51
...............vars
............activation_52
...............vars
............activation_53
...............vars
............activation_54
...............vars
............activation_55
...............vars
............activation_56
...............vars
............activation_57
...............vars
............activation_6
...............vars
............activation_7
...............vars
............activation_8
...............vars
............activation_9
...............vars
............add
...............vars
............add_1
...............vars
............add_10
...............vars
............add_11
...............vars
............add_12
...............vars
............add_13
...............vars
............add_14
...............vars
............add_15
...............vars
............add_16
...............vars
............add_17
...............vars
............add_18
...............vars
............add_19
...............vars
............add_2
...............vars
............add_20
...............vars
............add_21
...............vars
............add_22
...............vars
............add_23
...............vars
............add_24
...............vars
............add_25
...............vars
............add_3
...............vars
............add_4
...............vars
............add_5
...............vars
............add_6
...............vars
............add_7
...............vars
............add_8
...............vars
............add_9
...............vars
............batch_normalization
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_1
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_10
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_11
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_12
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_13
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_14
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_15
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_16
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_17
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_18
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_19
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_2
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_20
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_21
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_22
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_23
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_24
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_25
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_26
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_27
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_28
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_29
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_3
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_30
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_31
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_32
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_33
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_34
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_35
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_36
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_37
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_38
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_39
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_4
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_40
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_41
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_42
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_43
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_44
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_45
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_46
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_47
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_48
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_49
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_5
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_50
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_51
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_52
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_53
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_54
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_55
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_56
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_57
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_58
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_59
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_6
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_60
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_61
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_62
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_63
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_64
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_65
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_66
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_67
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_68
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_69
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_7
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_70
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_71
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_72
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_73
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_74
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_75
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_76
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_77
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_78
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_79
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_8
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_80
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_81
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_82
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_83
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_84
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_85
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_86
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_87
...............vars
..................0
..................1
..................2
..................3
............batch_normalization_9
...............vars
..................0
..................1
..................2
..................3
............conv2d
...............vars
..................0
............conv2d_1
...............vars
..................0
............conv2d_10
...............vars
..................0
............conv2d_100
...............vars
..................0
..................1
............conv2d_101
...............vars
..................0
..................1
............conv2d_102
...............vars
..................0
............conv2d_103
...............vars
..................0
............conv2d_104
...............vars
..................0
..................1
............conv2d_105
...............vars
..................0
..................1
............conv2d_106
...............vars
..................0
............conv2d_107
...............vars
..................0
............conv2d_108
...............vars
..................0
..................1
............conv2d_109
...............vars
..................0
..................1
............conv2d_11
...............vars
..................0
............conv2d_110
...............vars
..................0
............conv2d_111
...............vars
..................0
............conv2d_12
...............vars
..................0
............conv2d_13
...............vars
..................0
............conv2d_14
...............vars
..................0
............conv2d_15
...............vars
..................0
............conv2d_16
...............vars
..................0
..................1
............conv2d_17
...............vars
..................0
..................1
............conv2d_18
...............vars
..................0
............conv2d_19
...............vars
..................0
............conv2d_2
...............vars
..................0
............conv2d_20
...............vars
..................0
..................1
............conv2d_21
...............vars
..................0
..................1
............conv2d_22
...............vars
..................0
............conv2d_23
...............vars
..................0
............conv2d_24
...............vars
..................0
..................1
............conv2d_25
...............vars
..................0
..................1
............conv2d_26
...............vars
..................0
............conv2d_27
...............vars
..................0
............conv2d_28
...............vars
..................0
..................1
............conv2d_29
...............vars
..................0
..................1
............conv2d_3
...............vars
..................0
............conv2d_30
...............vars
..................0
............conv2d_31
...............vars
..................0
............conv2d_32
...............vars
..................0
..................1
............conv2d_33
...............vars
..................0
..................1
............conv2d_34
...............vars
..................0
............conv2d_35
...............vars
..................0
............conv2d_36
...............vars
..................0
..................1
............conv2d_37
...............vars
..................0
..................1
............conv2d_38
...............vars
..................0
............conv2d_39
...............vars
..................0
............conv2d_4
...............vars
..................0
............conv2d_40
...............vars
..................0
..................1
............conv2d_41
...............vars
..................0
..................1
............conv2d_42
...............vars
..................0
............conv2d_43
...............vars
..................0
............conv2d_44
...............vars
..................0
..................1
............conv2d_45
...............vars
..................0
..................1
............conv2d_46
...............vars
..................0
............conv2d_47
...............vars
..................0
............conv2d_48
...............vars
..................0
..................1
............conv2d_49
...............vars
..................0
..................1
............conv2d_5
...............vars
..................0
............conv2d_50
...............vars
..................0
............conv2d_51
...............vars
..................0
............conv2d_52
...............vars
..................0
..................1
............conv2d_53
...............vars
..................0
..................1
............conv2d_54
...............vars
..................0
............conv2d_55
...............vars
..................0
............conv2d_56
...............vars
..................0
..................1
............conv2d_57
...............vars
..................0
..................1
............conv2d_58
...............vars
..................0
............conv2d_59
...............vars
..................0
............conv2d_6
...............vars
..................0
............conv2d_60
...............vars
..................0
..................1
............conv2d_61
...............vars
..................0
..................1
............conv2d_62
...............vars
..................0
............conv2d_63
...............vars
..................0
............conv2d_64
...............vars
..................0
..................1
............conv2d_65
...............vars
..................0
..................1
............conv2d_66
...............vars
..................0
............conv2d_67
...............vars
..................0
............conv2d_68
...............vars
..................0
..................1
............conv2d_69
...............vars
..................0
..................1
............conv2d_7
...............vars
..................0
............conv2d_70
...............vars
..................0
............conv2d_71
...............vars
..................0
............conv2d_72
...............vars
..................0
..................1
............conv2d_73
...............vars
..................0
..................1
............conv2d_74
...............vars
..................0
............conv2d_75
...............vars
..................0
............conv2d_76
...............vars
..................0
..................1
............conv2d_77
...............vars
..................0
..................1
............conv2d_78
...............vars
..................0
............conv2d_79
...............vars
..................0
............conv2d_8
...............vars
..................0
............conv2d_80
...............vars
..................0
..................1
............conv2d_81
...............vars
..................0
..................1
............conv2d_82
...............vars
..................0
............conv2d_83
...............vars
..................0
............conv2d_84
...............vars
..................0
..................1
............conv2d_85
...............vars
..................0
..................1
............conv2d_86
...............vars
..................0
............conv2d_87
...............vars
..................0
............conv2d_88
...............vars
..................0
..................1
............conv2d_89
...............vars
..................0
..................1
............conv2d_9
...............vars
..................0
............conv2d_90
...............vars
..................0
............conv2d_91
...............vars
..................0
............conv2d_92
...............vars
..................0
..................1
............conv2d_93
...............vars
..................0
..................1
............conv2d_94
...............vars
..................0
............conv2d_95
...............vars
..................0
............conv2d_96
...............vars
..................0
..................1
............conv2d_97
...............vars
..................0
..................1
............conv2d_98
...............vars
..................0
............conv2d_99
...............vars
..................0
............depthwise_conv2d
...............vars
..................0
............depthwise_conv2d_1
...............vars
..................0
............depthwise_conv2d_10
...............vars
..................0
............depthwise_conv2d_11
...............vars
..................0
............depthwise_conv2d_12
...............vars
..................0
............depthwise_conv2d_13
...............vars
..................0
............depthwise_conv2d_14
...............vars
..................0
............depthwise_conv2d_15
...............vars
..................0
............depthwise_conv2d_16
...............vars
..................0
............depthwise_conv2d_17
...............vars
..................0
............depthwise_conv2d_18
...............vars
..................0
............depthwise_conv2d_19
...............vars
..................0
............depthwise_conv2d_2
...............vars
..................0
............depthwise_conv2d_20
...............vars
..................0
............depthwise_conv2d_21
...............vars
..................0
............depthwise_conv2d_22
...............vars
..................0
............depthwise_conv2d_23
...............vars
..................0
............depthwise_conv2d_3
...............vars
..................0
............depthwise_conv2d_4
...............vars
..................0
............depthwise_conv2d_5
...............vars
..................0
............depthwise_conv2d_6
...............vars
..................0
............depthwise_conv2d_7
...............vars
..................0
............depthwise_conv2d_8
...............vars
..................0
............depthwise_conv2d_9
...............vars
..................0
............dropout
...............vars
............dropout_1
...............vars
............dropout_10
...............vars
............dropout_11
...............vars
............dropout_12
...............vars
............dropout_13
...............vars
............dropout_14
...............vars
............dropout_15
...............vars
............dropout_16
...............vars
............dropout_17
...............vars
............dropout_18
...............vars
............dropout_19
...............vars
............dropout_2
...............vars
............dropout_20
...............vars
............dropout_21
...............vars
............dropout_22
...............vars
............dropout_23
...............vars
............dropout_24
...............vars
............dropout_25
...............vars
............dropout_3
...............vars
............dropout_4
...............vars
............dropout_5
...............vars
............dropout_6
...............vars
............dropout_7
...............vars
............dropout_8
...............vars
............dropout_9
...............vars
............global_average_pooling2d
...............vars
............global_average_pooling2d_1
...............vars
............global_average_pooling2d_10
...............vars
............global_average_pooling2d_11
...............vars
............global_average_pooling2d_12
...............vars
............global_average_pooling2d_13
...............vars
............global_average_pooling2d_14
...............vars
............global_average_pooling2d_15
...............vars
............global_average_pooling2d_16
...............vars
............global_average_pooling2d_17
...............vars
............global_average_pooling2d_18
...............vars
............global_average_pooling2d_19
...............vars
............global_average_pooling2d_2
...............vars
............global_average_pooling2d_20
...............vars
............global_average_pooling2d_21
...............vars
............global_average_pooling2d_22
...............vars
............global_average_pooling2d_23
...............vars
............global_average_pooling2d_3
...............vars
............global_average_pooling2d_4
...............vars
............global_average_pooling2d_5
...............vars
............global_average_pooling2d_6
...............vars
............global_average_pooling2d_7
...............vars
............global_average_pooling2d_8
...............vars
............global_average_pooling2d_9
...............vars
............input_layer
...............vars
............multiply
...............vars
............multiply_1
...............vars
............multiply_10
...............vars
............multiply_11
...............vars
............multiply_12
...............vars
............multiply_13
...............vars
............multiply_14
...............vars
............multiply_15
...............vars
............multiply_16
...............vars
............multiply_17
...............vars
............multiply_18
...............vars
............multiply_19
...............vars
............multiply_2
...............vars
............multiply_20
...............vars
............multiply_21
...............vars
............multiply_22
...............vars
............multiply_23
...............vars
............multiply_3
...............vars
............multiply_4
...............vars
............multiply_5
...............vars
............multiply_6
...............vars
............multiply_7
...............vars
............multiply_8
...............vars
............multiply_9
...............vars
............normalization
...............vars
............rescaling
...............vars
............reshape
...............vars
............reshape_1
...............vars
............reshape_10
...............vars
............reshape_11
...............vars
............reshape_12
...............vars
............reshape_13
...............vars
............reshape_14
...............vars
............reshape_15
...............vars
............reshape_16
...............vars
............reshape_17
...............vars
............reshape_18
...............vars
............reshape_19
...............vars
............reshape_2
...............vars
............reshape_20
...............vars
............reshape_21
...............vars
............reshape_22
...............vars
............reshape_23
...............vars
............reshape_3
...............vars
............reshape_4
...............vars
............reshape_5
...............vars
............reshape_6
...............vars
............reshape_7
...............vars
............reshape_8
...............vars
............reshape_9
...............vars
.........vars
......global_average_pooling2d
.........vars
...metrics
......auc
.........vars
............0
............1
............2
............3
......mean
.........vars
............0
............1
......mean_metric_wrapper
.........vars
............0
............1
......precision
.........vars
............0
............1
......recall
.........vars
............0
............1
...optimizer
......vars
.........0
.........1
.........2
.........3
.........4
...vars
Keras model archive saving:
File Name                                             Modified             Size
metadata.json                                  2023-03-25 03:32:56           64
config.json                                    2023-03-25 03:32:56       200166
variables.h5                                   2023-03-25 03:32:58     52793232
time: 3.2 s (started: 2023-03-25 03:32:56 +00:00)


# Save the Object Detection model to a file in the current working directory
Pkl_Filename = "Pickle_ObjectDetection_UNetMobileNet_Model.pkl"  

with open(Pkl_Filename, 'wb') as file:  
    pickle.dump(model9, file)

# # Load the DL Model back from the file
# with open(Pkl_Filename, 'rb') as file:  
#     Pickle_ObjectDetection_UNetMobileNet_Model = pickle.load(file)
    
# Pickle_ObjectDetection_UNetMobileNet_Model

Keras weights file (<HDF5 file "variables.h5" (mode r+)>) saving:
...layers
......batch_normalization
.........vars
............0
............1
............2
............3
......batch_normalization_1
.........vars
............0
............1
............2
............3
......batch_normalization_10
.........vars
............0
............1
............2
............3
......batch_normalization_11
.........vars
............0
............1
............2
............3
......batch_normalization_12
.........vars
............0
............1
............2
............3
......batch_normalization_13
.........vars
............0
............1
............2
............3
......batch_normalization_14
.........vars
............0
............1
............2
............3
......batch_normalization_15
.........vars
............0
............1
............2
............3
......batch_normalization_16
.........vars
............0
............1
............2
............3
......batch_normalization_17
.........vars
............0
............1
............2
............3
......batch_normalization_18
.........vars
............0
............1
............2
............3
......batch_normalization_19
.........vars
............0
............1
............2
............3
......batch_normalization_2
.........vars
............0
............1
............2
............3
......batch_normalization_20
.........vars
............0
............1
............2
............3
......batch_normalization_21
.........vars
............0
............1
............2
............3
......batch_normalization_22
.........vars
............0
............1
............2
............3
......batch_normalization_23
.........vars
............0
............1
............2
............3
......batch_normalization_24
.........vars
............0
............1
............2
............3
......batch_normalization_25
.........vars
............0
............1
............2
............3
......batch_normalization_26
.........vars
............0
............1
............2
............3
......batch_normalization_3
.........vars
............0
............1
............2
............3
......batch_normalization_4
.........vars
............0
............1
............2
............3
......batch_normalization_5
.........vars
............0
............1
............2
............3
......batch_normalization_6
.........vars
............0
............1
............2
............3
......batch_normalization_7
.........vars
............0
............1
............2
............3
......batch_normalization_8
.........vars
............0
............1
............2
............3
......batch_normalization_9
.........vars
............0
............1
............2
............3
......concatenate
.........vars
......concatenate_1
.........vars
......concatenate_2
.........vars
......concatenate_3
.........vars
......conv2d
.........vars
............0
......conv2d_1
.........vars
............0
......conv2d_10
.........vars
............0
......conv2d_11
.........vars
............0
......conv2d_12
.........vars
............0
......conv2d_13
.........vars
............0
......conv2d_14
.........vars
............0
............1
......conv2d_2
.........vars
............0
......conv2d_3
.........vars
............0
......conv2d_4
.........vars
............0
......conv2d_5
.........vars
............0
......conv2d_6
.........vars
............0
......conv2d_7
.........vars
............0
......conv2d_8
.........vars
............0
......conv2d_9
.........vars
............0
......depthwise_conv2d
.........vars
............0
......depthwise_conv2d_1
.........vars
............0
......depthwise_conv2d_10
.........vars
............0
......depthwise_conv2d_11
.........vars
............0
......depthwise_conv2d_12
.........vars
............0
......depthwise_conv2d_2
.........vars
............0
......depthwise_conv2d_3
.........vars
............0
......depthwise_conv2d_4
.........vars
............0
......depthwise_conv2d_5
.........vars
............0
......depthwise_conv2d_6
.........vars
............0
......depthwise_conv2d_7
.........vars
............0
......depthwise_conv2d_8
.........vars
............0
......depthwise_conv2d_9
.........vars
............0
......input_layer
.........vars
......re_lu
.........vars
......re_lu_1
.........vars
......re_lu_10
.........vars
......re_lu_11
.........vars
......re_lu_12
.........vars
......re_lu_13
.........vars
......re_lu_14
.........vars
......re_lu_15
.........vars
......re_lu_16
.........vars
......re_lu_17
.........vars
......re_lu_18
.........vars
......re_lu_19
.........vars
......re_lu_2
.........vars
......re_lu_20
.........vars
......re_lu_21
.........vars
......re_lu_22
.........vars
......re_lu_23
.........vars
......re_lu_24
.........vars
......re_lu_25
.........vars
......re_lu_26
.........vars
......re_lu_3
.........vars
......re_lu_4
.........vars
......re_lu_5
.........vars
......re_lu_6
.........vars
......re_lu_7
.........vars
......re_lu_8
.........vars
......re_lu_9
.........vars
......reshape
.........vars
......up_sampling2d
.........vars
......up_sampling2d_1
.........vars
......up_sampling2d_2
.........vars
......up_sampling2d_3
.........vars
......up_sampling2d_4
.........vars
......zero_padding2d
.........vars
......zero_padding2d_1
.........vars
......zero_padding2d_2
.........vars
......zero_padding2d_3
.........vars
...metrics
......auc
.........vars
............0
............1
............2
............3
......mean
.........vars
............0
............1
......mean_io_u
.........vars
............0
......mean_metric_wrapper
.........vars
............0
............1
......mean_metric_wrapper_1
.........vars
............0
............1
......mean_metric_wrapper_2
.........vars
............0
............1
......precision
.........vars
............0
............1
......recall
.........vars
............0
............1
...optimizer
......vars
.........0
.........1
.........10
.........100
.........101
.........102
.........103
.........104
.........105
.........106
.........107
.........108
.........109
.........11
.........110
.........111
.........112
.........113
.........114
.........115
.........116
.........117
.........118
.........119
.........12
.........120
.........121
.........122
.........123
.........124
.........125
.........126
.........127
.........128
.........129
.........13
.........130
.........131
.........132
.........133
.........134
.........135
.........136
.........137
.........138
.........139
.........14
.........140
.........141
.........142
.........143
.........144
.........145
.........146
.........147
.........148
.........149
.........15
.........150
.........151
.........152
.........153
.........154
.........155
.........156
.........157
.........158
.........159
.........16
.........160
.........161
.........162
.........163
.........164
.........165
.........166
.........17
.........18
.........19
.........2
.........20
.........21
.........22
.........23
.........24
.........25
.........26
.........27
.........28
.........29
.........3
.........30
.........31
.........32
.........33
.........34
.........35
.........36
.........37
.........38
.........39
.........4
.........40
.........41
.........42
.........43
.........44
.........45
.........46
.........47
.........48
.........49
.........5
.........50
.........51
.........52
.........53
.........54
.........55
.........56
.........57
.........58
.........59
.........6
.........60
.........61
.........62
.........63
.........64
.........65
.........66
.........67
.........68
.........69
.........7
.........70
.........71
.........72
.........73
.........74
.........75
.........76
.........77
.........78
.........79
.........8
.........80
.........81
.........82
.........83
.........84
.........85
.........86
.........87
.........88
.........89
.........9
.........90
.........91
.........92
.........93
.........94
.........95
.........96
.........97
.........98
.........99
...vars
Keras model archive saving:
File Name                                             Modified             Size
metadata.json                                  2023-03-25 03:32:59           64
config.json                                    2023-03-25 03:32:59        52863
variables.h5                                   2023-03-25 03:33:00     38921048
time: 1.02 s (started: 2023-03-25 03:32:59 +00:00)


!lscpu |grep 'Model name'

Model name:                      Intel(R) Xeon(R) CPU @ 2.00GHz
time: 340 ms (started: 2023-03-25 03:33:00 +00:00)


# Memory that we can use
# max memory used = 8GB
!free -h --si | awk  '/Mem:/{print $2}'

12G
time: 323 ms (started: 2023-03-25 03:33:00 +00:00)


# if it had turbo boost it would've shown Min and Max MHz also but it is only showing current frequency 
# this means it always operates at shown frequency
!lscpu | grep "MHz"

CPU MHz:                         2000.146
time: 327 ms (started: 2023-03-25 03:33:01 +00:00)


# hard disk space that we can use
!df -h / | awk '{print $4}'

Avail
51G
time: 324 ms (started: 2023-03-25 03:33:01 +00:00)


# no.of cores each processor is having 
!lscpu | grep 'Core(s) per socket:'

Core(s) per socket:              1
time: 347 ms (started: 2023-03-25 03:33:01 +00:00)


# no.of threads each core is having
!lscpu | grep 'Thread(s) per core'

Thread(s) per core:              2
time: 322 ms (started: 2023-03-25 03:33:02 +00:00)


# no.of sockets i.e available slots for physical processors
!lscpu | grep 'Socket(s):'

Socket(s):                       1
time: 331 ms (started: 2023-03-25 03:33:02 +00:00)


!lscpu | grep "L3 cache"

L3 cache:                        38.5 MiB
time: 317 ms (started: 2023-03-25 03:33:02 +00:00)


# use this command to see GPU activity while doing Deep Learning tasks, 
# for this command 'nvidia-smi' and for above one to work, go to 
# 'Runtime > change runtime type > Hardware Accelerator > GPU'
!nvidia-smi

Sat Mar 25 03:33:03 2023       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.85.12    Driver Version: 525.85.12    CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla T4            Off  | 00000000:00:04.0 Off |                    0 |
| N/A   71C    P0    31W /  70W |  14553MiB / 15360MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+
time: 425 ms (started: 2023-03-25 03:33:03 +00:00)


# GPU count and name
!nvidia-smi -L

GPU 0: Tesla T4 (UUID: GPU-1db530b7-d509-2cd0-bd55-24d2dd4e3db7)
time: 335 ms (started: 2023-03-25 03:33:03 +00:00)


# Timestamp
!rm /etc/localtime
!ln -s /usr/share/zoneinfo/Asia/Bangkok /etc/localtime
!date

Sat 25 Mar 2023 10:33:04 AM +07
time: 980 ms (started: 2023-03-25 03:33:03 +00:00)


# NB end time
b = time.time()
# Total NB Runtime
# print(f"Total NB Run Time (min.): {(b-a)/60}")

time: 559 µs (started: 2023-03-25 03:42:37 +00:00)

	patientId	class
13157	84d7f055-3374-4d68-9390-cf3b1418355c	No Lung Opacity / Not Normal
20611	bc4ec257-ecf8-469b-9750-c5bd190f6987	Lung Opacity
8627	60f16c3a-9a51-4c68-b5c4-2ce898c08370	No Lung Opacity / Not Normal
4731	40ef3f33-d990-4a08-b652-ed4e02e84159	Lung Opacity
15854	9a05ffb1-0988-4f71-89a3-2eac51deb62e	Lung Opacity
3083	3601ea72-9ad8-4638-aca9-6f22d945bd01	No Lung Opacity / Not Normal
29053	025d196b-cfa4-44a5-9e31-0f73dc3b318f	Lung Opacity
20196	b9b46200-1045-4c7a-b790-a9f6cdd031e3	Lung Opacity
8099	5c9364c7-51da-490b-82bd-e7078e9c21f2	Normal
14222	8d11ed1c-a608-47d4-ad04-5d8733c24bdd	No Lung Opacity / Not Normal

	patientId	x	y	width	height	Target
25939	e7c89fed-c3fc-4207-9009-d9105713f094	NaN	NaN	NaN	NaN	0
15293	95ae7ef0-fb02-4663-a7dd-a74b81514dab	NaN	NaN	NaN	NaN	0
550	088265f7-0350-4c51-a7fc-99f1ddc89ae1	NaN	NaN	NaN	NaN	0
6804	51daa411-4fb0-42f0-aa15-46c332656c04	NaN	NaN	NaN	NaN	0
17294	a595fe39-ce96-43dd-ac73-89d49124b966	537.0	224.0	303.0	692.0	1
3820	3ae783bc-9fd3-4e66-b9cd-101a43f122bd	605.0	495.0	183.0	196.0	1
16807	a1f6801d-e44b-4d7c-bef0-066328f9bd94	NaN	NaN	NaN	NaN	0
29570	1d3e4775-9d2c-4e5f-81de-ce29b540a9f7	NaN	NaN	NaN	NaN	0
13627	88a033e5-b48c-4f0c-aa0a-3c0ae353cc0d	NaN	NaN	NaN	NaN	0
23448	d316817a-0747-4dfb-be81-fbf857439d42	595.0	402.0	284.0	297.0	1

	patientId	x	y	width	height	Target
5	00436515-870c-4b36-a041-de91049b9ab4	562.0	152.0	256.0	453.0	1
9	00704310-78a8-4b38-8475-49f4573b2dbb	695.0	575.0	162.0	137.0	1

	patientId	x	y	width	height	Target
4	00436515-870c-4b36-a041-de91049b9ab4	264.0	152.0	213.0	379.0	1
5	00436515-870c-4b36-a041-de91049b9ab4	562.0	152.0	256.0	453.0	1

	patientId	x	y	width	height	class
20102	9d244e11-444b-4add-84b5-52e277742ca6	NaN	NaN	NaN	NaN	Normal
8171	4b831076-3931-46db-ac95-0e11df232af8	NaN	NaN	NaN	NaN	No Lung Opacity / Not Normal
15346	7db0ba60-d789-4bdf-afb6-298dc68fbdd8	NaN	NaN	NaN	NaN	No Lung Opacity / Not Normal
6876	4212e538-53db-4327-8028-b429230b5030	NaN	NaN	NaN	NaN	No Lung Opacity / Not Normal
17465	8b5c52fa-61d9-4d89-acaf-20a1b63a309d	NaN	NaN	NaN	NaN	Normal

	x	y	width	height	patientId	Target
Total	20672.000000	20672.000000	20672.000000	20672.000000	0.0	0.0
Percent	68.389188	68.389188	68.389188	68.389188	0.0	0.0

	patientId	x	y	width	height	Target	class
10	00704310-78a8-4b38-8475-49f4573b2dbb	323.0	577.0	160.0	104.0	1	Lung Opacity
11	00704310-78a8-4b38-8475-49f4573b2dbb	323.0	577.0	160.0	104.0	1	Lung Opacity
12	00704310-78a8-4b38-8475-49f4573b2dbb	695.0	575.0	162.0	137.0	1	Lung Opacity
13	00704310-78a8-4b38-8475-49f4573b2dbb	695.0	575.0	162.0	137.0	1	Lung Opacity

	Exams	Target	Class	Entries
0	1	0	No Lung Opacity / Not Normal	11821
1	1	0	Normal	8851
2	1	1	Lung Opacity	2614
3	4	1	Lung Opacity	3266
4	9	1	Lung Opacity	119
5	16	1	Lung Opacity	13

	patientId	x	y	width	height	Target	class	Modality	PatientAge	PatientSex	BodyPartExamined	ViewPosition	ConversionType	Rows	Columns	PixelSpacing
12037	67cbdff5-178b-4da2-9a0e-7c7ab1bdcaf1	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal	CR	37	F	CHEST	AP	WSD	1024	1024	0.168
18705	93351a14-30a7-494c-a02b-7f2d72c724c8	612.0	161.0	228.0	552.0	1	Lung Opacity	CR	21	F	CHEST	AP	WSD	1024	1024	0.139
4985	38fe4a2d-b910-4eb6-9465-64a353d66c02	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal	CR	34	M	CHEST	AP	WSD	1024	1024	0.139
9902	57e70848-0cce-456d-90b1-e3f0a206d8c1	NaN	NaN	NaN	NaN	0	Normal	CR	29	F	CHEST	PA	WSD	1024	1024	0.168
14171	75ecaed3-c912-4459-9095-ec6674f524e3	NaN	NaN	NaN	NaN	0	Normal	CR	8	M	CHEST	PA	WSD	1024	1024	0.139
27663	c4e517d8-5970-4a03-bcf4-180d6fc65548	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal	CR	70	M	CHEST	PA	WSD	1024	1024	0.143
1256	0b6acd03-8392-49a6-bf52-4543fb678a25	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal	CR	64	F	CHEST	AP	WSD	1024	1024	0.139
31093	de7f40b2-741b-4345-881c-856526ec76dc	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal	CR	28	M	CHEST	AP	WSD	1024	1024	0.139
32334	e7a73b81-7634-46b1-bd54-6465b85c2009	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal	CR	29	M	CHEST	PA	WSD	1024	1024	0.143
1794	14a7fbc6-6661-4382-882f-b2aa317cadc0	705.0	380.0	159.0	211.0	1	Lung Opacity	CR	42	M	CHEST	AP	WSD	1024	1024	0.168

	patientId	PredictionString
888	133d3fcb-f893-430a-9ee0-07a720e4b19e	0.5 0 0 100 100
1160	1afa2510-edc7-4ecc-ae32-743403295afe	0.5 0 0 100 100
1640	21425cab-c511-4129-961c-ce5fc01715b1	0.5 0 0 100 100
2894	c0662eac-8a07-4ba1-9c8a-e4d15c3ac067	0.5 0 0 100 100
1611	20d9afc7-a8aa-414d-9e44-fce8775b0dbb	0.5 0 0 100 100
1573	205f6d99-0870-499f-9644-92b98d985d62	0.5 0 0 100 100
903	13659019-299a-47bc-ac85-25500b625fde	0.5 0 0 100 100
400	05aa63ed-15bb-43eb-ba2a-08ebe2029c12	0.5 0 0 100 100
372	0540925d-6075-4bb6-9007-020b53c08958	0.5 0 0 100 100
260	03a75f93-bdfb-4e5a-89f3-548b04fdbb5e	0.5 0 0 100 100

	patientId	Modality	PatientAge	PatientSex	BodyPartExamined	ViewPosition	ConversionType	Rows	Columns	PixelSpacing
1051	19b00444-7cea-4504-a557-bd207e023377	CR	51	M	CHEST	PA	WSD	1024	1024	0.143
945	13bd80c3-a3a0-4a2a-83c2-4b298a1f47a0	CR	22	F	CHEST	AP	WSD	1024	1024	0.139
1253	1c3f2cc0-4c64-43bf-828f-8d52f7c5a442	CR	8	M	CHEST	PA	WSD	1024	1024	0.139
1981	25e0c806-3608-4b47-a1e9-f6f0c7787eff	CR	77	M	CHEST	PA	WSD	1024	1024	0.194
1693	21f523b0-7213-4d04-94d7-4ca690ecb69d	CR	52	F	CHEST	PA	WSD	1024	1024	0.143
1087	19fe1b8d-e93e-4810-b8cb-19c59141c90c	CR	41	M	CHEST	PA	WSD	1024	1024	0.194
2121	27d3a0fe-120f-488e-8cb1-4dd73afa755c	CR	64	F	CHEST	AP	WSD	1024	1024	0.139
698	11149595-46a2-48b8-91d3-b9ebc4da8d88	CR	6	F	CHEST	AP	WSD	1024	1024	0.168
1311	1d0de7b7-c3db-424c-bfd1-fe63c121b668	CR	52	M	CHEST	PA	WSD	1024	1024	0.143
1862	243c5673-c107-4cc0-bcd5-6d5bc8a1a8dd	CR	24	M	CHEST	AP	WSD	1024	1024	0.168

	patientId	x	y	width	height	Target
7227	554a744a-36f2-4e93-bda2-7ac51b583331	NaN	NaN	NaN	NaN	0
11932	7ae347da-a4b0-4e3c-979e-33a509e40649	NaN	NaN	NaN	NaN	0
22602	cc269df0-a264-4980-8d99-97a50322cc9c	NaN	NaN	NaN	NaN	0
20350	ba8f8cb0-5aed-4242-b117-8516f4c7b51b	NaN	NaN	NaN	NaN	0
18207	ac58847c-9f1a-4248-85a1-58ac5d10781a	NaN	NaN	NaN	NaN	0
25519	e49d1683-b3d9-4a75-9b30-779201943f1a	132.0	214.0	224.0	382.0	1
17370	a62c5825-e9e5-4e8b-8cc5-8a09aeb1f85a	NaN	NaN	NaN	NaN	0
10277	6de46d7a-2863-463d-a46a-fd85af10783b	NaN	NaN	NaN	NaN	0
12481	7f3c3430-4f1a-4793-b2d3-7ec2b637e2f5	NaN	NaN	NaN	NaN	0
24861	df215f73-bde0-4d55-9e55-8d49ff3ee2f4	NaN	NaN	NaN	NaN	0

	patientId	class
16095	9c27b40f-9d27-4b00-be43-b7953199783a	Normal
11729	79310bce-abe6-4e7d-a285-2601cfeeca5e	Lung Opacity
11271	7581fe5e-1611-4563-a327-e140d4327ce8	No Lung Opacity / Not Normal
2474	321d8cc5-0fc3-43de-b12f-851515e800d5	Lung Opacity
29499	1b52079c-b9ef-42d3-9e9a-17aea3bc6031	Normal
7434	56ea3620-70fb-403f-8cf6-86ece871c47e	Normal
20406	bae7b652-311f-4a84-a031-653424887f35	Lung Opacity
28882	ff33da14-0bd6-4fe4-9675-ef704bc33b84	Lung Opacity
18635	af42a7f3-092c-4d10-b5a2-5806a1a8aea6	Lung Opacity
29099	03e4827c-7338-4de3-9ac6-8831ba5637e9	Lung Opacity

	patientId	x	y	width	height	Target	class
0	0004cfab-14fd-4e49-80ba-63a80b6bddd6	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal
1	00313ee0-9eaa-42f4-b0ab-c148ed3241cd	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal
2	00322d4d-1c29-4943-afc9-b6754be640eb	NaN	NaN	NaN	NaN	0	No Lung Opacity / Not Normal
3	003d8fa0-6bf1-40ed-b54c-ac657f8495c5	NaN	NaN	NaN	NaN	0	Normal
4	00436515-870c-4b36-a041-de91049b9ab4	264.0	152.0	213.0	379.0	1	Lung Opacity

	loss	accuracy	precision	recall	AUC	val_loss	val_accuracy	val_precision	val_recall	val_AUC	lr
0	49849.207031	0.345860	0.345860	0.345860	0.509395	2102.844727	0.543257	0.543257	0.543257	0.657443	0.010000
1	1260.365234	0.481529	0.481529	0.481529	0.610810	504.229034	0.521628	0.521628	0.521628	0.640368	0.008913
2	331.954803	0.541401	0.541401	0.541401	0.656732	263.955078	0.484733	0.484733	0.484733	0.613522	0.007943
3	166.439743	0.553503	0.553503	0.553503	0.665832	177.886536	0.449109	0.449109	0.449109	0.588518	0.007079
4	128.254028	0.535350	0.535350	0.535350	0.653528	125.866425	0.552163	0.552163	0.552163	0.665690	0.006310
5	142.282593	0.500637	0.500637	0.500637	0.626685	186.914368	0.508906	0.508906	0.508906	0.634579	0.005623
6	166.783493	0.502866	0.502866	0.502866	0.628209	136.788132	0.520356	0.520356	0.520356	0.641367	0.005012
7	177.295303	0.468153	0.468153	0.468153	0.602736	125.740799	0.516539	0.516539	0.516539	0.643371	0.004467
8	102.626175	0.555732	0.555732	0.555732	0.670205	62.574825	0.600509	0.600509	0.600509	0.701198	0.003981
9	59.015663	0.578981	0.578981	0.578981	0.688172	44.141380	0.585242	0.585242	0.585242	0.691110	0.003548
10	44.161621	0.584713	0.584713	0.584713	0.694559	133.152267	0.423664	0.423664	0.423664	0.567541	0.003162
11	48.231647	0.568471	0.568471	0.568471	0.678942	32.183952	0.606870	0.606870	0.606870	0.707680	0.002818
12	31.273994	0.614968	0.614968	0.614968	0.716060	52.562325	0.461832	0.461832	0.461832	0.601037	0.002512
13	39.337997	0.576433	0.576433	0.576433	0.687051	46.630562	0.472010	0.472010	0.472010	0.607126	0.002239
14	24.599930	0.621338	0.621338	0.621338	0.726842	33.347065	0.496183	0.496183	0.496183	0.630332	0.001995
15	27.794769	0.593631	0.593631	0.593631	0.702036	38.333553	0.541985	0.541985	0.541985	0.664989	0.001778
16	24.218998	0.618471	0.618471	0.618471	0.724987	28.169601	0.580153	0.580153	0.580153	0.692684	0.001585
17	21.310295	0.623885	0.623885	0.623885	0.727739	38.209267	0.524173	0.524173	0.524173	0.647295	0.001413
18	20.337744	0.621019	0.621217	0.621019	0.727458	19.307865	0.610687	0.610191	0.609415	0.718978	0.001259
19	11.301879	0.692994	0.693214	0.692994	0.787367	19.267494	0.594148	0.594148	0.594148	0.713973	0.001122
20	12.813560	0.660510	0.660510	0.660510	0.760905	30.374329	0.500000	0.500000	0.500000	0.632151	0.001000
21	12.917532	0.671656	0.671656	0.671656	0.767853	14.907911	0.618321	0.618321	0.618321	0.726134	0.000891
22	8.417212	0.716560	0.716560	0.716560	0.809543	16.803469	0.548346	0.548346	0.548346	0.676538	0.000794
23	7.885828	0.723885	0.724116	0.723885	0.808765	13.890759	0.632316	0.632316	0.632316	0.736802	0.000708
24	8.972069	0.693631	0.693631	0.693631	0.787995	13.613379	0.632316	0.632316	0.632316	0.737761	0.000631
25	6.507399	0.743949	0.743949	0.743949	0.827535	14.919806	0.580153	0.580153	0.580153	0.695284	0.000562
26	6.110028	0.749363	0.749283	0.749045	0.834999	13.737605	0.575064	0.575064	0.575064	0.691471	0.000501
27	7.854130	0.700000	0.700000	0.700000	0.793561	11.835890	0.619593	0.620382	0.619593	0.733748	0.000447
28	6.039867	0.736943	0.737177	0.736943	0.825587	14.447656	0.580153	0.580153	0.580153	0.706047	0.000398
29	5.431669	0.755414	0.755655	0.755414	0.840363	12.668351	0.615776	0.615776	0.615776	0.730130	0.000355

	loss	accuracy	precision	recall	AUC	val_loss	val_accuracy	val_precision	val_recall	val_AUC	lr
count	30.000000	30.000000	30.000000	30.000000	30.000000	30.000000	30.000000	30.000000	30.000000	30.000000	30.000000
mean	1758.658075	0.611062	0.611096	0.611051	0.718082	144.310147	0.549873	0.549883	0.549830	0.669829	0.002968
std	9085.805497	0.098043	0.098078	0.098027	0.081092	383.986755	0.058230	0.058245	0.058185	0.047862	0.002763
min	5.431669	0.345860	0.345860	0.345860	0.509395	11.835890	0.423664	0.423664	0.423664	0.567541	0.000355
25%	9.554521	0.554060	0.554060	0.554060	0.666925	15.390721	0.510814	0.510814	0.510814	0.636026	0.000819
50%	26.197350	0.616720	0.616720	0.616720	0.720523	35.778166	0.550254	0.550254	0.550254	0.671114	0.001887
75%	121.847065	0.693471	0.693527	0.693471	0.787838	125.835018	0.598919	0.598919	0.598919	0.707272	0.004345
max	49849.207031	0.755414	0.755655	0.755414	0.840363	2102.844727	0.632316	0.632316	0.632316	0.737761	0.010000

	param	score mean	score s.d.	rank
0	{'lr': 0.01, 'batch_size': 32}	0.345649	0.044398	7
1	{'lr': 0.1, 'batch_size': 32}	0.392778	0.080756	2
2	{'lr': 0.001, 'batch_size': 32}	0.413139	0.053356	1
3	{'lr': 0.01, 'batch_size': 64}	0.348193	0.045510	6
4	{'lr': 0.1, 'batch_size': 64}	0.348696	0.045456	5
5	{'lr': 0.001, 'batch_size': 64}	0.384896	0.061763	3
6	{'lr': 0.01, 'batch_size': 128}	0.306402	0.039810	9
7	{'lr': 0.1, 'batch_size': 128}	0.315597	0.036713	8
8	{'lr': 0.001, 'batch_size': 128}	0.364221	0.075290	4

	Model	Train Accuracy	Test Accuracy	Precision	Recall	F1 Score	AUC
6	Model-7: EfficientNetV2B3	0.772542	0.693405	0.694492	0.702172	0.696146	0.869599
5	Model-6: ResNet50	0.799032	0.695187	0.703094	0.694715	0.698141	0.879281
4	Model-5: VGG16	0.739175	0.644682	0.640697	0.654945	0.644738	0.829839
1	Model-2: CNN-1	0.653846	0.647059	0.646612	0.650587	0.642724	0.827212
7	Model-8: CNN-1 (Tuned)	0.649007	0.642305	0.646541	0.643015	0.642730	0.827652
3	Model-4: MobileNetV2	0.709628	0.638146	0.635131	0.642407	0.630174	0.826473
0	Model-1: Base Model	0.726439	0.632799	0.653712	0.623473	0.628346	0.745432
2	Model-3: CNN-2	0.609272	0.612597	0.615194	0.611098	0.611342	0.800113

	patientId	x	y	width	height	Target
15386	966db49a-1aa9-4d43-bedd-f2f58e08ceee	NaN	NaN	NaN	NaN	0
26908	ef7c58ec-8b59-4406-952e-9d54c34e6276	693.0	585.0	209.0	156.0	1
4780	415419bd-51a7-4708-8cb9-1adcba6921b1	571.0	599.0	185.0	214.0	1
18346	ad4d16e5-a851-4327-b9fe-5977da445669	189.0	513.0	190.0	286.0	1
8033	5c0e0efc-38f9-4335-8283-77502ffd5738	285.0	484.0	125.0	254.0	1
23462	d346540b-05a3-43f3-9d55-858aaa915c22	NaN	NaN	NaN	NaN	0
20236	b9f8516d-e115-4f48-a43f-79d4ba74840e	NaN	NaN	NaN	NaN	0
13893	8aaaa6ce-64d0-4a03-a01e-0db5007ad818	NaN	NaN	NaN	NaN	0
4386	3e9e02d2-673c-4a8d-bdfe-35515693ce1f	NaN	NaN	NaN	NaN	0
24404	db5399f8-e0cb-4208-8694-a7136f9ac574	378.0	310.0	111.0	130.0	1

	patientId	class
14261	8d7363fb-fb76-4052-855c-913cc592b5b1	No Lung Opacity / Not Normal
17240	a51d69b5-64a1-4fbb-a7f1-7f8a4e1fc953	No Lung Opacity / Not Normal
14871	92307af5-1f76-473f-beb1-45d94fe66a7c	No Lung Opacity / Not Normal
1157	0e0cfa1a-c097-47bb-bf4b-6b9362940705	Lung Opacity
589	08d8e57a-4835-445b-8c79-8dd31ab17819	Lung Opacity
26170	e9d01262-68be-48c0-b278-3ada6fae8933	No Lung Opacity / Not Normal
24103	d8ba6755-6d47-4c47-875a-a6aa0c55e3bf	No Lung Opacity / Not Normal
27029	f06b985f-1424-44e3-9c57-530a93012786	No Lung Opacity / Not Normal
18488	ae505ed5-51b5-44f7-8145-6cfe7785b756	No Lung Opacity / Not Normal
6810	51e7f8d7-2904-4651-a3e3-be6e8d351cfc	No Lung Opacity / Not Normal

	loss	accuracy	precision	recall	f1_score	AUC	dice_coef	mean_iou	val_loss	val_accuracy	val_precision	val_recall	val_f1_score	val_AUC	val_dice_coef	val_mean_iou	lr
0	0.873197	0.010868	0.084013	0.395739	0.126867	0.637974	0.126803	0.482411	0.892857	0.894000	0.000000	0.000000	0.000000	0.500000	0.107143	0.490212	0.010000
1	0.829828	0.002179	0.124392	0.337062	0.170121	0.621653	0.170172	0.496595	0.979243	0.692655	0.018706	0.006851	0.002934	0.498830	0.020757	0.491088	0.008913
2	0.873937	0.005102	0.115633	0.218151	0.126011	0.582772	0.126063	0.503115	0.894266	0.816101	0.118607	0.061240	0.054695	0.530534	0.105734	0.501404	0.007943
3	0.835370	0.011373	0.159364	0.224731	0.164614	0.585988	0.164630	0.508094	0.892857	0.717793	0.000000	0.000000	0.000000	0.499933	0.107143	0.490199	0.007079
4	0.811297	0.010563	0.141711	0.427022	0.188742	0.666250	0.188703	0.504360	0.868512	0.551313	0.129959	0.054350	0.043559	0.525841	0.131488	0.503556	0.006310
5	0.764091	0.010510	0.193321	0.406328	0.235984	0.670801	0.235909	0.530927	0.854778	0.199139	0.120063	0.318881	0.145190	0.638923	0.145222	0.520345	0.005623
6	0.753094	0.007334	0.188393	0.501742	0.246796	0.713851	0.246906	0.534509	0.885350	0.287893	0.129915	0.071641	0.043124	0.531898	0.114650	0.508696	0.005012
7	0.795259	0.009540	0.163634	0.419871	0.204765	0.674962	0.204741	0.520922	0.827328	0.372662	0.141103	0.168588	0.104185	0.576443	0.172672	0.522158	0.004467
8	0.789715	0.010895	0.155484	0.500142	0.210257	0.705946	0.210285	0.518489	0.883102	0.284160	0.124636	0.071120	0.044852	0.534395	0.116898	0.505144	0.003981
9	0.762398	0.027756	0.197805	0.403541	0.237549	0.671072	0.237602	0.533927	0.875840	0.096381	0.122065	0.238705	0.124068	0.605483	0.124160	0.521805	0.003548
10	0.763177	0.031755	0.180056	0.500206	0.236771	0.714963	0.236823	0.534610	0.839299	0.234588	0.134643	0.322750	0.160736	0.645600	0.160701	0.525620	0.003162
11	0.754861	0.041215	0.203452	0.441273	0.245227	0.687548	0.245140	0.535787	0.819982	0.115460	0.134022	0.501566	0.180082	0.720821	0.180018	0.525682	0.002818
12	0.734501	0.043769	0.205917	0.535268	0.268105	0.735236	0.268064	0.547741	0.814592	0.063536	0.136261	0.511095	0.185381	0.726490	0.185408	0.527142	0.002512
13	0.738922	0.057996	0.220372	0.434928	0.261163	0.694687	0.261078	0.552818	0.766575	0.077633	0.149668	0.440803	0.179802	0.699132	0.233425	0.535229	0.002239
14	0.760386	0.056295	0.217953	0.384498	0.239599	0.669458	0.239615	0.547160	0.789460	0.062473	0.188104	0.351076	0.193318	0.667466	0.210540	0.550061	0.001995
15	0.776548	0.052854	0.218330	0.345808	0.223488	0.654247	0.223453	0.546185	0.805354	0.034014	0.255561	0.201765	0.159254	0.608038	0.194646	0.539845	0.001778
16	0.725869	0.031035	0.243225	0.417976	0.278011	0.684445	0.278007	0.550738	0.804369	0.012131	0.122187	0.692896	0.195694	0.800852	0.195631	0.512362	0.001585
17	0.725130	0.022906	0.227526	0.471703	0.274869	0.707313	0.274870	0.552992	0.793428	0.014908	0.133938	0.680553	0.206498	0.801561	0.206572	0.523818	0.001413
18	0.708254	0.017020	0.247726	0.461963	0.291858	0.708131	0.291746	0.559742	0.796403	0.005434	0.131567	0.671502	0.203647	0.800055	0.203597	0.527230	0.001259
19	0.689957	0.034425	0.263171	0.478436	0.310033	0.718542	0.310042	0.567457	0.783389	0.022972	0.141231	0.676550	0.216583	0.804297	0.216611	0.531380	0.001122

	loss	accuracy	precision	recall	f1_score	AUC	dice_coef	mean_iou	val_loss	val_accuracy	val_precision	val_recall	val_f1_score	val_AUC	val_dice_coef	val_mean_iou	lr
count	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000	20.000000
mean	0.773290	0.024769	0.187574	0.415319	0.227041	0.675292	0.227033	0.531429	0.843349	0.277762	0.121612	0.302097	0.122180	0.635830	0.156651	0.517649	0.004138
std	0.050792	0.018028	0.047099	0.084145	0.051106	0.042040	0.051095	0.023106	0.052728	0.295317	0.058505	0.251929	0.077646	0.111400	0.052728	0.016816	0.002704
min	0.689957	0.002179	0.084013	0.218151	0.126011	0.582772	0.126063	0.482411	0.766575	0.005434	0.000000	0.000000	0.000000	0.498830	0.020757	0.490199	0.001122
25%	0.737817	0.010550	0.158394	0.392929	0.200760	0.663250	0.200731	0.515890	0.802378	0.055358	0.121565	0.068650	0.044529	0.531557	0.116336	0.504747	0.001941
50%	0.762788	0.019963	0.195563	0.423446	0.237160	0.679704	0.237212	0.534559	0.833314	0.157300	0.130763	0.278793	0.152222	0.623480	0.166686	0.521982	0.003355
75%	0.799269	0.036123	0.218841	0.473386	0.262898	0.707518	0.262824	0.548490	0.883664	0.417324	0.137471	0.503949	0.187365	0.722238	0.197622	0.527164	0.005795
max	0.873937	0.057996	0.263171	0.535268	0.310033	0.735236	0.310042	0.567457	0.979243	0.894000	0.255561	0.692896	0.216583	0.804297	0.233425	0.550061	0.010000

	Model	Train Accuracy	Test Accuracy	Precision	Recall	F1 Score	AUC	Dice Coef	MeanIoU
0	Model-9: UNet (MobileNet)	0.024769	0.024219	0.218881	0.602574	0.283223	0.767537	0.283200	0.556895
1	Model-10: UNet (ResNet50)	0.008163	0.002801	0.173479	0.632654	0.239430	0.768630	0.239386	0.533566

Problem Statement: Project Specs¶

Milestone-1¶

EXPLORATORY DATA ANALYSIS¶

Load the data¶

Merge the data¶

Explore the DICOM data¶

Dicom metadata¶

Plot DICOM images¶

Add meta information from the dicom data¶

Train dataset¶

Test dataset¶

Modality Attribute¶

BodyPartExamined Attribute¶

ViewPosition Attribute¶

Train dataset¶

Test dataset¶

ConversionType Attribute¶

Rows and Columns¶

PatientAge Attribute¶

Train dataset¶

Test dataset¶

PatientGender Attribute¶

Train dataset¶

Test dataset¶

PixelSpacing¶

Train dataset¶

Test dataset¶

MODELING: CLASSIFICATION¶

1. Prepare the data.¶

Model-1: Base Model¶

Model-2: CNN-1¶

Model-3: CNN-2¶

Milestone-2¶

Model-4: MobileNetV2¶

Model-5: VGG16¶

Model-6: ResNet50¶

Model-7: EfficientNetV2B3¶

Model-8: CNN-1 (Tuned)¶

Model Comparison - Classification¶

MODELING: OBJECT DETECTION¶

1. Prepare the data.¶

Performance Metrics¶

Model-9: UNet (MobileNet)¶

Model-10: UNet (ResNet50)¶

Model Comparison - Object Detection¶

Pickle the model¶

Project Summary:¶

System Resources Used¶

References:¶